# Altyst: full text

> Underwrite any property in seconds. Altyst reads the OM, rent roll, and T-12 and builds a live, auditable real-estate model with Excel and PDF outputs.

This file is the complete public reference content of https://altyst.ai as plain text: the
underwriting answer library, the long-form guides, and the product and pricing facts.
The index version, with links only, is at https://altyst.ai/llms.txt.

Everything here is published on a page a visitor can read. Nothing is asserted that the
site does not state. Illustrative figures are labelled as illustrative and describe no
real property or transaction.

# Product facts

Altyst is real-estate underwriting software. You upload the documents a deal already
comes with, an offering memorandum, a rent roll, a T-12, a debt quote, a lease schedule, and
it produces a complete, editable underwriting model plus the Excel workbook and PDF
documents an investment committee expects.

- AI reads documents and proposes values, each shown with its source and confidence for
  human review. It never performs the calculation.
- A tested, deterministic engine computes every financial result in exact decimal
  arithmetic. The same inputs always produce the same result.
- Every input traces back to the document it was read from, carrying the exact text and, where the extractor can pin it, the page. Every output carries the formula that produced it.
- Asset classes: multifamily, single-family and short-term rental, office, medical office,
  retail, industrial, flex, warehouse, self-storage, mixed-use, hotel and hospitality, and
  land or development including ground-up.
- Inputs: PDF, Excel, CSV, Word, plain text, and photographs of a page. Scanned pages are
  read with OCR.
- Outputs: an Excel workbook on live formulas, a one-page investment memo PDF, an editable
  PowerPoint deck, a deck PDF, a lender package, and an offering summary.
- Optional layers, off by default: an after-tax view using rates the user supplies, and a
  GP and LP distribution waterfall.
- Support: support@altyst.ai

# Pricing

- Individual: $12 a month, 5 deals a month, $4 per additional deal.
- Professional: $24 a month, 15 deals a month, $3 per additional deal.
- Team: $99 a month, 5 seats, 75 deals a month, $3 per additional deal.
- Enterprise: contact sales.
- Each plan has a monthly ceiling on new deals, included plus additional: 25 on Individual, 60 on Professional, and 200 on Team. Enterprise ceilings are set per agreement.

Every plan includes the full model, every asset class, and the Excel and PDF outputs. Plans
bill monthly and renew until cancelled. Modeling, recompute, and exports are never metered;
AI document reading and property research draw on an AI allowance carried by each deal,
included or additional, so one deal cannot use up another deal's. One unit of usage is one
deal creation. Unused deal credits roll over.

# Answers

## What is a T-12 in real estate?
Source: https://altyst.ai/answers/what-is-a-t12
Updated: 2026-08-06

Answer: A T-12 is a trailing twelve month operating statement: every income and expense line a property actually produced over the last twelve months, almost always presented as twelve monthly columns with an annual total. Buyers ask for it because it is the one document in a deal package that reports what happened rather than what the seller projects. Underwriting starts by normalizing it, which means restating the seller's expenses at what they will cost under your ownership.

### What a T-12 contains

Rows are general ledger lines, columns are the twelve months. Revenue runs from gross potential rent down through loss to lease, vacancy, concessions and bad debt, then adds other income to reach effective gross income. Below that sit the operating expenses: real estate taxes, insurance, utilities, repairs and maintenance, contract services, payroll, the management fee, general and administrative, and marketing. The statement usually keeps going past net operating income into capital items, debt service and owner distributions.

The monthly presentation is the useful part. An annual total hides a tax bill that landed in one month, a utility line that doubled in July, and a repair line that was flat for eleven months and then absorbed a roof.
T-12 shape, illustrative:
- Gross potential rent | 4,871,000
- Vacancy and credit loss | (341,000)
- Other income | 198,000
- Effective gross income | 4,728,000
- Operating expenses | (1,708,000)
- Net operating income | 3,020,000

### The lines that are wrong most often

Capital work booked as repairs is the classic. A full roof replacement sitting in R and M depresses net operating income, which sounds like it favors the buyer, until you realize the seller has already told you the roof was replaced and is now marketing on a pro forma that excludes it. Read the repairs line item by item, not in total.

The management fee is the second. An owner who self-manages often books a fee well below market, or none at all. If you will pay a third party four percent of effective gross income, that is the number in your model regardless of what the T-12 says.

Then there is anything non-recurring: an insurance rebate, a one-time legal settlement, a tax appeal refund, a burst pipe. Each one belongs in the trailing statement and none of them belongs in a forward year.

### Why the last three months often matter more than the twelve

Practitioners pull a T-3 and a T-1 alongside the T-12 and annualize them. If the T-3 annualized is materially above the T-12, something changed for the better recently: a lease-up finished, concessions burned off, a renovation program started producing premiums. If it is below, look for a tenant that left or an expense that reset.

Insurance and real estate taxes are the two lines where the trailing number is most likely to be irrelevant to you. Taxes usually reassess on sale in states that permit it, so the seller's basis is not yours. Insurance in coastal and wildfire markets has repriced faster than any trailing statement can show. Both of these should come from a quote or a calculation, never from the T-12.

### Normalizing before it becomes a pro forma

Normalization is the step between the trailing statement and year one of the model. It is where most of the real underwriting judgment lives, and it is almost never shown in the offering memorandum. The short version is below; the long version, adjustment by adjustment, is in the guide on what a T-12 hides.
- Real estate taxes at your expected assessed value and millage, not the seller's
- Insurance at a quoted premium for your coverage, not the legacy policy
- Management at the market fee you will actually pay
- Payroll adjusted if you will staff the property differently
- Replacement reserves added per unit or per square foot
- One-time items stripped out entirely

### The T-12 has to tie to the rent roll

Rent roll rents annualized, less vacancy and concessions, should land close to the T-12 revenue. When they disagree by more than a rounding error, the gap is a fact about the deal and not a data problem. Common causes: concessions the rent roll shows as asking rent, units offline, a tenant paying below the stated lease rate, or other income captured in one document and not the other.

Reconciling those two documents by hand is a slow and unglamorous hour, which is why it often gets skipped on a first pass. Altyst reads the rent roll and the T-12 together and surfaces the places they disagree, with the page each value came from, so the review is on the exceptions rather than on the re-keying.

Related questions

Q: What does T-12 stand for?
A: Trailing twelve months. A T-12 is an operating statement covering the most recent twelve months of actual property income and expense, usually shown as twelve monthly columns.

Q: Is a T-12 the same as a profit and loss statement?
A: Effectively yes. A T-12 is a property level profit and loss statement presented as twelve monthly columns rather than a single annual figure, which is what makes seasonality and one-time items visible.

Q: What is a T-3 in real estate?
A: A T-3 is the trailing three months of operating results. Investors annualize it and compare it to the T-12 to see whether recent performance is running above or below the trailing year.

Q: Why does the T-12 net operating income differ from the offering memorandum?
A: The offering memorandum usually shows a pro forma net operating income, which applies the seller's assumptions about future rents, occupancy and expenses. The T-12 shows actual results. The difference between the two is the seller's growth story, and it is the buyer's job to test it.

---

## How do you read a rent roll?
Source: https://altyst.ai/answers/how-to-read-a-rent-roll
Updated: 2026-08-06

Answer: A rent roll is a snapshot of what is leased right now: unit by unit for residential, tenant by tenant for commercial, showing who occupies the space, at what rent, and until when. Read it for three things. What is actually being collected today, how far that sits from market, and when each lease rolls.

### The columns that carry the information

A residential rent roll gives you unit number, floor plan, square footage, tenant, lease start and end, in-place rent, a market or asking rent, security deposit, and a status flag. A commercial rent roll adds the recovery structure, the expense stop or base year, options, and often a percentage rent clause.

Everything else on the page is administrative. The four fields that drive a model are square footage, in-place rent, lease end date, and status.

### Loss to lease is a gap, not a promise

Loss to lease is the difference between what the units would produce at market rent and what the in-place leases actually produce. A large number is the entire pitch of most value-add deals, and it deserves more suspicion than it usually gets.

Two questions decide whether it is real. First, is the market rent column a market rent or the property manager's aspiration? Test it against your own comps, not against the offering memorandum's comps. Second, how fast can you capture it? On twelve month leases every unit rolls once inside year one, but they roll on a staggered schedule, so a unit that turns in month nine pays the new rent for three months of that year. Even a perfect execution collects roughly half the annual uplift in year one. Capture also costs money: turn cost, downtime, and often a concession.

### Physical, leased, and economic occupancy are three different numbers

Physical occupancy counts units with somebody living in them. Leased occupancy counts units under a signed lease including future move-ins. Economic occupancy is collected revenue divided by gross potential rent, and it is the only one that shows up in cash flow.

The spread between them is where the story hides. A property at ninety five percent physical and eighty seven percent economic is telling you about concessions, bad debt, employee units, model units, or units held offline for renovation. None of those are visible in the headline occupancy figure a broker leads with.

### A rent roll is also an expiration schedule

Sort by lease end date and you have the rollover profile. For multifamily this mostly matters for seasonality and for how quickly a renovation program can move. For office, retail and industrial it is the deal. A single year holding forty percent of the income is a cliff, and it needs a renewal assumption, a downtime assumption, and the capital to re-tenant if the assumption is wrong.

Look at credit concentration at the same time. One tenant at thirty percent of income means you are underwriting that tenant as much as the building.

### Where a rent roll misleads

Asking rent is not effective rent. Two months free on a twelve month lease means the tenant pays ten months of rent over a twelve month term, which is a seventeen percent discount that no column on the page displays. Renewals in progress are sometimes shown at the pending rent rather than the current one. Month to month tenants may sit at a premium that will not survive a real market test. And a rent roll dated more than a few weeks before your review is not current.

The discipline that catches all of it is the same one: reconcile the rent roll to the T-12. Annualized rent roll revenue, less vacancy and concessions, should land near trailing actual revenue.

Altyst extracts the rent roll and the trailing statement together, ties one against the other, and flags each disagreement with the document and page it came from. Every extracted value is flagged for your review, carrying its source and a confidence signal, and you can override any of them.

Related questions

Q: What is loss to lease?
A: Loss to lease is the difference between gross potential rent at market rates and the rent the in-place leases actually produce. It represents rent that could be captured as leases roll, net of the turn cost, downtime and concessions required to capture it.

Q: What is economic occupancy?
A: Economic occupancy is collected rental revenue divided by gross potential rent. It differs from physical occupancy because it also absorbs concessions, bad debt, non-revenue units and units held offline.

Q: What is the difference between a rent roll and a T-12?
A: A rent roll is a point-in-time snapshot of current leases and rents. A T-12 is twelve months of actual income and expense. The rent roll tells you what should be collected going forward; the T-12 tells you what was collected historically.

Q: How current does a rent roll need to be?
A: Within a few weeks of your review. Leases roll continuously, so a rent roll more than a month old can misstate occupancy, in-place rents and the near-term expiration schedule.

---

## How do you calculate net operating income (NOI)?
Source: https://altyst.ai/answers/how-to-calculate-noi
Updated: 2026-08-06

Answer: Net operating income is effective gross income minus operating expenses. Effective gross income is gross potential rent less vacancy, concessions and credit loss, plus other income. The arithmetic is trivial; the difficulty is deciding which lines count as operating expenses, because debt service, capital expenditure, tenant improvements, leasing commissions, depreciation and income taxes are all excluded by definition.

### The revenue stack

Start at gross potential rent, which is every unit or suite at market rent as if fully occupied and paying. Subtract loss to lease to get to contract rent, then subtract vacancy, concessions and bad debt. Add other income: parking, storage, laundry, pet fees, utility reimbursement, late fees, and in commercial deals the expense recoveries. The result is effective gross income.

Order matters less than consistency. What matters is that every deduction appears exactly once and that the model uses the same stack in year one that it uses in year ten.
Revenue to NOI, illustrative:
- Gross potential rent | 4,871,000
- Less vacancy and credit loss | (341,000)
- Plus other income | 198,000
- Effective gross income | 4,728,000
- Less operating expenses | (1,708,000)
- Net operating income | 3,020,000

### What counts as an operating expense

Real estate taxes, property insurance, utilities not billed back, repairs and maintenance, contract services such as landscaping and trash, on-site payroll, the property management fee, general and administrative, marketing and leasing costs that are not commissions.

The test is whether the cost recurs in the ordinary course of running the property. A quarterly HVAC service contract does. Replacing the HVAC system does not.

### What is excluded, and why people get it wrong anyway

Debt service is excluded because net operating income is a property-level measure, independent of how the buyer financed it. That is the whole reason cap rates are comparable across deals. Capital expenditure, tenant improvements and leasing commissions are excluded because they are investments in the asset rather than the cost of operating it. Depreciation and income taxes are excluded because they are attributes of the owner, not the building. Owner distributions and partnership costs never belonged there at all.

The exclusion people fight over is replacement reserves. The US convention treats reserves as below the line, so a reported net operating income usually does not carry them. Lenders and appraisers frequently deduct a reserve anyway, at a per-unit or per-square-foot rate, before they size a loan or apply a cap rate. Both conventions are defensible. What is not defensible is buying on a no-reserve net operating income and then selling into an appraisal that deducts one.
- Excluded: debt service, principal, interest
- Excluded: capital expenditure, tenant improvements, leasing commissions
- Excluded: depreciation, amortization, income taxes
- Excluded: owner distributions, entity level costs
- Convention-dependent: replacement reserves

### Why two analysts get two different NOIs on the same building

Because normalization is a set of choices. Management fee at the seller's rate or at market. Taxes at the current assessment or at the reassessed value after sale. Insurance at the legacy premium or at a fresh quote. Payroll at the seller's staffing or yours. General vacancy at physical occupancy or at a stabilized underwriting standard. Every one of those is a judgment, and each one moves the number.

This is why a net operating income figure without its assumptions is not information. It is a headline.

### Whose NOI goes in the cap rate

A cap rate is net operating income divided by value, but three different net operating incomes get used: trailing, forward twelve months, and stabilized. The trailing number prices what the asset does today. The forward number prices the next year under your ownership and is what most buyers actually transact on. The stabilized number prices a future that requires capital and time to reach, and quoting a cap rate on it without discounting for that risk is how a deal gets overpaid for.

Altyst builds the whole stack explicitly, with rent and expenses on separate growth curves, a per-year growth schedule where a flat rate will not do, and custom lines that carry their own growth rate, so a tax line capped by statute does not have to drift with everything else. Every figure exposes the formula behind it and the inputs it depends on.

Related questions

Q: What is the formula for NOI?
A: Net operating income equals effective gross income minus operating expenses. Effective gross income equals gross potential rent, less vacancy, concessions and credit loss, plus other income and recoveries.

Q: Is debt service included in NOI?
A: No. Net operating income is measured before debt service, which is what makes it comparable across properties regardless of how each buyer financed the purchase.

Q: Are capital expenditures included in NOI?
A: No. Capital expenditure, tenant improvements and leasing commissions are investments in the asset and sit below net operating income, not inside it.

Q: Should replacement reserves be deducted before NOI?
A: By US convention reserves sit below the line, so reported net operating income usually excludes them. Lenders and appraisers commonly deduct a per-unit or per-square-foot reserve anyway when sizing debt or applying a cap rate, so it is worth knowing which convention a quoted figure uses.

---

## How do lenders size a commercial real estate loan?
Source: https://altyst.ai/answers/how-lenders-size-a-commercial-real-estate-loan
Updated: 2026-08-06

Answer: A lender runs three independent tests and lends the smallest result. A loan-to-value cap limits the loan to a percentage of appraised value. A minimum debt service coverage ratio limits it to the debt the property's net operating income can service, usually tested at a stressed rate. A debt yield floor limits it to net operating income divided by a required yield. Which of the three binds tells you more about the market than any single ratio does.

### Loan to value

The simplest test: loan divided by appraised value, capped at whatever the lender's program allows. It is the least informative of the three because it depends entirely on an appraisal, and an appraisal depends on a cap rate, which is the thing most likely to have moved since the last comparable sale.

### Debt service coverage ratio

Coverage is net operating income divided by annual debt service. A 1.25x minimum means the property must produce twenty five percent more income than the loan payment consumes. To size from it, invert the calculation: maximum debt service is net operating income divided by the required coverage, and maximum loan is that debt service divided by the mortgage constant.

The mortgage constant is annual debt service per dollar of loan, and it is where the rate and the amortization schedule enter. At a six percent rate on a thirty year amortization the constant is roughly 0.0719, so every dollar of annual debt service supports about fourteen dollars of loan. Shorten the amortization to twenty five years and the constant rises, the loan shrinks, and nothing about the property changed.

Most lenders test coverage at a stress rate rather than the contract rate, and some test it on a stress constant with an assumed amortization even when the loan is interest only. Always ask which.
Three tests, illustrative:
- Loan to value | 65% of $47.0M | $30.6M
- DSCR 1.25x | NOI $3.02M, constant 7.19% | $33.6M
- Debt yield 9.0% | NOI $3.02M | $33.6M
- Binding constraint | The smallest of the three | $30.6M

### Debt yield

Debt yield is net operating income divided by the loan amount. A nine percent floor means the lender will lend at most about eleven dollars for every dollar of net operating income. It became the discipline of choice after the last cycle because it contains no rate assumption, no amortization assumption and no appraisal. It is the only one of the three tests that cannot be loosened by a friendly assumption.

That is exactly why it binds in hot markets. When values run ahead of income, loan-to-value and coverage both stay comfortable while debt yield does not.

### Which constraint binds, and what it tells you

In a low rate environment coverage is easy, and because low rates push values up relative to income, the loan-to-value cap is loose too, so the debt yield floor is usually what binds. As rates rise, the constant rises with them and coverage starts binding first, which is why proceeds fell across the board when rates repriced even though nothing happened to the buildings. In a wide-cap market, where value is low relative to income, loan-to-value takes over.

The practical consequence is that a sponsor who models a fixed leverage percentage is modeling a fiction. Leverage is an output of whichever test binds, and it moves during diligence.

### Interest only, refinance risk, and the second loan

An interest-only period raises coverage and cash-on-cash for as long as it lasts, and it hides nothing about the loan itself, only about the year the amortization starts. Model the year it burns off, not the average.

Refinance risk is the same three tests applied at a future date to a future net operating income at a future rate. If the takeout loan is smaller than the balance, the gap is equity, and it is due whether the plan worked or not.

Altyst sizes debt on a loan-to-value cap, a minimum coverage test at a stress rate, or a debt yield floor, whichever binds, and reports coverage and debt yield by year. It supports an interest-only period, amortization, senior and mezzanine tranches, and a bridge to refinance with cash-out.

Related questions

Q: What is a debt service coverage ratio?
A: The debt service coverage ratio is net operating income divided by annual debt service. A 1.25x ratio means the property produces twenty five percent more income than the loan payment requires. Lenders set a minimum and size the loan so the ratio is met, often testing it at a stressed interest rate.

Q: What is debt yield?
A: Debt yield is net operating income divided by the loan amount, expressed as a percentage. It measures the return a lender would earn if it took the property back, and unlike loan-to-value or coverage it contains no interest rate, amortization or appraisal assumption.

Q: What is a mortgage constant?
A: The mortgage constant is annual debt service per dollar of loan principal, given a rate and an amortization schedule. At a six percent rate amortizing over thirty years it is roughly 7.19 percent, so one dollar of annual debt service supports about fourteen dollars of loan.

Q: Why did loan proceeds fall when interest rates rose?
A: Because a higher rate raises the mortgage constant, so each dollar of net operating income supports less debt under the coverage test. The property's income did not change; the amount of loan that income can service did.

---

## Cap rate, cash-on-cash, or IRR: which return metric should you use?
Source: https://altyst.ai/answers/cap-rate-vs-cash-on-cash-vs-irr
Updated: 2026-08-06

Answer: They answer three different questions, so the choice is not between them. A cap rate prices the asset unlevered at a single moment: net operating income divided by value. Cash-on-cash measures one year of levered cash flow against the equity invested. Internal rate of return measures the entire hold including the sale, weighted by when each dollar arrives. Read together with the equity multiple they describe a deal; read alone, each one is easy to game.

### Cap rate: the price of the income

Net operating income divided by price. It ignores financing entirely, which is its strength: two buyers with different debt still see the same cap rate on the same building, so it works as a pricing comparison across deals and across markets.

Its weakness is that it is a single year and it ignores capital. A building that needs eight million dollars of deferred maintenance trades at a cap rate that says nothing about the eight million. Going-in, exit and stabilized cap rates are three different numbers, and a quoted cap rate without which one it is means very little.

### Cash-on-cash: does it feed itself

Levered cash flow after debt service, divided by the equity invested, for a given year. This is the metric that tells an investor whether the deal distributes cash while they hold it, which matters enormously to some investors and not at all to others.

It says nothing about the sale and nothing about the years you are not looking at. An interest-only period flatters it for exactly as long as the interest-only period lasts.

### IRR: the whole hold, time weighted

The discount rate at which the deal's cash flows net to zero. It is the closest thing to a complete answer, and it is also the easiest to manipulate, because it rewards speed. A shorter hold at the same multiple produces a higher internal rate of return, so a deal can be engineered to look better by assuming an earlier exit rather than a better outcome.

It is also dominated by the exit assumption. In a five year hold the sale is usually the majority of the value, which means the internal rate of return is substantially a statement about the exit cap rate. Change the exit cap by fifty basis points and watch what happens.

Levered and unlevered internal rates of return answer different questions again. The unlevered figure measures the real estate. The levered figure measures the real estate plus the financing decision.

### Equity multiple keeps IRR honest

Total distributions divided by total equity. It has no opinion about time, which is precisely why it belongs next to a metric that is all about time. A 2.19x equity multiple over five years with no interim distributions is roughly a seventeen percent internal rate of return; the same 2.19x over eight years is closer to ten percent. Neither number is wrong and neither is sufficient.
Same multiple, different holds, illustrative:
- 2.19x | 5 years | 17%
- 2.19x | 8 years | 10%
- 1.60x | 3 years | 17%

### Yield on cost and the development spread

For anything with a construction or renovation budget, the metric that matters is yield on cost: stabilized net operating income divided by total cost including land, hard costs, soft costs and carry. Compare it to the cap rate the finished asset would trade at. The gap between the two is the development spread, and it is the compensation for taking the execution risk.

If the spread is thin, no internal rate of return presentation makes the deal safe. It just makes it leveraged.

### How to quote them without misleading anyone

The habit that prevents most of the damage is a labelling discipline. Never quote a cap rate without saying whether it is going-in, exit or stabilized. Never show a cash-on-cash without the year it belongs to, since year one and the year the interest-only burns off are different deals. Never show an internal rate of return without the equity multiple beside it and the exit cap that produced it, because that assumption is usually carrying the number. And anything with a construction or renovation budget gets a yield on cost quoted against the market cap rate for the finished asset.

Altyst reports levered and unlevered internal rate of return, equity multiple, cash-on-cash by year and on average, yield on cost, and development profit and margin, alongside base, upside and downside cases. The internal rate of return solver is decimal-exact and guards against the spurious roots a spreadsheet function can return on a sign-alternating cash flow.

Related questions

Q: What is a cap rate?
A: A capitalization rate is net operating income divided by property value or price. It expresses the unlevered first-year yield on the asset and is used to compare pricing across properties, since it excludes financing.

Q: What is the difference between cash-on-cash return and IRR?
A: Cash-on-cash return measures one year of levered cash flow against the equity invested and ignores the sale. Internal rate of return measures every cash flow across the entire hold, including the sale proceeds, weighted by when each one occurs.

Q: Why can IRR be misleading?
A: Because it rewards early cash. A shorter hold at the same equity multiple produces a higher internal rate of return, and in a typical five year hold the figure is dominated by the assumed exit cap rate rather than by operations.

Q: What is yield on cost?
A: Yield on cost is stabilized net operating income divided by total project cost including land, hard costs, soft costs and carry. Compared against the market cap rate for the finished asset, the difference is the development spread.

---

## How does a real estate distribution waterfall work?
Source: https://altyst.ai/answers/how-a-real-estate-waterfall-works
Updated: 2026-08-06

Answer: A distribution waterfall is the order in which cash from a property is split between the limited partners who funded it and the general partner who runs it. The common sequence pays a preferred return on limited partner capital first, then returns that capital, then splits whatever remains on terms that favor the general partner. That disproportionate share of the remainder is the promote, and it is the sponsor's actual compensation for performance.

### The four steps in a common structure

Cash available for distribution flows down a set of tiers, and each tier is filled before the next one receives anything. A typical structure runs: preferred return to the limited partners, return of limited partner capital, a general partner catch-up, then a residual split.

Everything interesting about a waterfall is in the details of those tiers, not in the headline split. Two deals both described as eighty twenty can distribute very differently.
A common four-tier structure, illustrative:
- 8% preferred return | 100% | 0%
- Return of capital | 100% | 0%
- GP catch-up | 0% | 100%
- Residual | 80% | 20%

### The preferred return is a hurdle, not a coupon

A preferred return is a priority claim on cash, not a promise of payment. If the property does not distribute, the pref does not get paid. What happens next depends on whether it is cumulative, in which case the shortfall accrues and must be made up before the general partner sees anything, or non-cumulative, in which case it simply does not happen that year.

Whether it compounds matters more than the rate. An eight percent pref that accrues and compounds annually on unreturned capital is a materially larger claim over a five year hold than the same rate simple, and the difference lands entirely on the general partner's promote.

### The catch-up, explained without the jargon

After the limited partners have received their preferred return and their capital back, a catch-up tier sends the next distributions to the general partner, often one hundred percent of them, until the general partner has received its promote percentage of the profits distributed so far.

The purpose is arithmetic rather than generosity. Without a catch-up, an eighty twenty split applied only to the residual means the general partner ends up with less than twenty percent of total profit, because the preferred tier paid limited partners alone. With a full catch-up, the general partner reaches a true twenty percent of profit.

A partial catch-up, say fifty percent, lands somewhere between the two. It is a common negotiated middle and it is often the single most valuable term in the document.

### Tiered promotes: IRR hurdles versus multiple hurdles

Many structures add tiers that increase the general partner's share as performance improves. A frequent shape is eighty twenty to a twelve percent internal rate of return, seventy thirty to eighteen percent, sixty forty above that.

Hurdles measured on internal rate of return are path dependent. Because the metric is time weighted, a general partner can cross a hurdle by selling sooner rather than by producing more, which is a real conflict when the sponsor controls the exit timing. Hurdles measured on an equity multiple have no such property, which is why sophisticated limited partners often ask for both tests, requiring a deal to clear an internal rate of return hurdle and a multiple hurdle before the higher split applies.

### Deal by deal, whole fund, and the clawback

An American waterfall computes the promote deal by deal, so a sponsor can earn promote on a winner while another investment is still under water. A European waterfall computes it across the whole fund, so limited partners get all of their capital and pref back before any promote is paid. The first is friendlier to sponsors and the second to investors, and the gap between them is usually bridged by a clawback provision requiring the general partner to return promote that later proves to have been premature.

A clawback is only as good as the entity behind it. Ask what secures it.

### Modeling it

The waterfall sits after the property model, not inside it. Property cash flow and sale proceeds are computed first, then the split is applied to the resulting distributions. That separation is what lets the same deal be tested under different partnership terms without touching the underwriting.

Altyst supports a waterfall with a compounding preferred return, return of capital, an optional general partner catch-up, a residual promote, and optional multi-tier internal rate of return hurdles. It is off by default, because a deal with no partnership structure should not be shown one.

Related questions

Q: What is a preferred return in real estate?
A: A preferred return is a priority claim on distributions paid to limited partners before the general partner shares in profits, usually quoted as an annual percentage of unreturned capital. It is a hurdle rather than a promised payment: if the property distributes nothing, nothing is paid, and whether the shortfall accrues depends on whether the pref is cumulative.

Q: What is a promote in real estate?
A: The promote, also called carried interest, is the general partner's disproportionate share of profits above the preferred return and return of capital. In an eighty twenty structure the general partner receives twenty percent of residual profit while having contributed a much smaller share of the equity.

Q: What is a GP catch-up?
A: A catch-up tier directs distributions to the general partner, often entirely, until the general partner has received its stated promote percentage of all profits distributed so far. Without it, an eighty twenty split applied only after a preferred return leaves the general partner with less than twenty percent of total profit.

Q: What is the difference between an American and a European waterfall?
A: An American waterfall calculates promote deal by deal, so a sponsor can earn promote on one investment while another is still under water. A European waterfall calculates it across the whole fund, returning all limited partner capital and preferred return before any promote is paid. Clawback provisions exist to correct promote paid too early under the American structure.

---

## How do you model lease rollover, TI and LC in office and retail?
Source: https://altyst.ai/answers/lease-rollover-ti-and-lc
Updated: 2026-08-06

Answer: Model each tenant on its own expiration, not on a portfolio average. At every roll date decide the market rent, a renewal probability, and the capital each outcome requires: renewal terms with a lower tenant improvement allowance and a shorter downtime, or a new lease with a higher allowance, a full leasing commission, and months of vacancy before rent restarts. The blended result of those two outcomes is the cash flow. The capital is usually what decides the deal.

### The roll itself

Three inputs per expiration. The market rent you expect at that future date, which is today's market rent grown forward and not today's contract rent. The renewal probability, which is a judgment about the tenant and the space. And the term of whatever replaces the expiring lease.

A weighted blend of renew and re-tenant is the standard approach and it is fine for portfolio cash flow. It is not fine for a single large tenant, where the two outcomes are so different that an average describes neither. Model that one discretely and look at both branches.

### Tenant improvements and leasing commissions

Tenant improvement allowances are quoted per square foot and paid near lease commencement, not spread over the term. New-lease allowances run materially above renewal allowances because a renewing tenant does not need the space rebuilt. Leasing commissions are a percentage of lease value and are usually split between the tenant broker and the listing broker, with renewals commissioned at a lower rate or not at all.

Both are real cash, both land in a year, and both sit below net operating income. A model that reports net operating income growing smoothly while quietly consuming several million dollars of tenant improvement and commission capital in year three is not lying, but it is not communicating either.
Roll assumptions, illustrative:
- Tenant improvement, per sq ft | $15 | $60
- Leasing commission | 2% | 5%
- Downtime | 0 months | 9 months
- Free rent | 1 month | 4 months

### Downtime and free rent are not the same thing

Downtime is the vacancy between one lease ending and the next one paying. Free rent is abatement inside a signed lease. They both reduce collected rent and they behave differently: downtime also stops expense recoveries, while an abated tenant is often still reimbursing operating expenses.

Both should be explicit. Folding them into a single occupancy percentage destroys the timing, and timing is the entire point of modeling a roll.

### Recoveries, expense stops and base years

The recovery structure determines who absorbs expense growth. Triple net passes essentially all operating expense through. A base year or expense stop structure passes through only the growth above a fixed level, which means the landlord keeps the base and the tenant takes the increases. Full service gross leaves the landlord with everything.

The trap is a base year that resets on a new lease. Re-tenanting a suite in a base year building hands the new tenant a fresh stop at current expense levels, which quietly removes several years of accumulated recovery income. It shows up nowhere in the rent comparison and everywhere in the cash flow.

Retail adds percentage rent, where the tenant pays a share of sales above a breakpoint. It is genuine income and it is also the most volatile line in a retail model.

### Rollover concentration is a capital problem

Sort the expiration schedule and look for the year that holds a large share of income. That year needs a renewal assumption, a downtime assumption, and the balance sheet to fund the tenant improvement and commission bill if the assumption is wrong. Lenders look at this before they look at the going-in coverage.

Altyst rolls each tenant on its own expiration with tenant improvements, leasing commissions, downtime, and a renewal probability, and it re-rolls each generation across a long hold rather than stopping after the first cycle. Base-year expense stops reset at rollover the way they do in a real lease, percentage rent is computed over a natural or stated breakpoint, and free rent is booked as abatement with recoveries still billing. Two limits worth knowing before you rely on it: expirations resolve to annual periods rather than to a month, and renewal options and month-to-month tenancies are not modeled as such, so a lease that lives or dies on an option date is one to check by hand.

Related questions

Q: What are TI and LC in commercial real estate?
A: TI is a tenant improvement allowance, the capital a landlord contributes to build out a space, quoted per square foot and paid near lease commencement. LC is the leasing commission paid to brokers on a new or renewed lease, usually a percentage of total lease value.

Q: Why are renewal TI allowances lower than new lease allowances?
A: A renewing tenant is already in the space and does not need it rebuilt, so the allowance covers refresh work rather than a full build-out. Renewal leasing commissions are lower for the same reason: less brokerage work is involved.

Q: What is an expense stop?
A: An expense stop is a fixed level of operating expense the landlord absorbs, with the tenant reimbursing growth above it. A base year structure sets that level at the actual expenses of the lease's first year, which means a new lease resets the stop to current expense levels.

Q: What is downtime in a lease rollover model?
A: Downtime is the number of months a space sits vacant between one lease expiring and the next lease beginning to pay rent. It is separate from free rent, which is abatement inside a signed lease during which expense recoveries often continue.

---

## How should you read an offering memorandum?
Source: https://altyst.ai/answers/how-to-read-an-offering-memorandum
Updated: 2026-08-06

Answer: An offering memorandum is the seller's marketing document. It is usually the best available summary of the asset and the least reliable set of forward numbers in the package, because the pro forma inside it is an argument rather than an underwriting. Read the exhibits first, build your own year one from the rent roll and the T-12, and treat the narrative as a source of questions rather than answers.

### What is inside one

An executive summary and the investment highlights, a property description, a market and submarket overview, rent and sale comparables, a financial summary with a trailing statement and a pro forma, and exhibits: the rent roll, the T-12, sometimes a tax bill, a capital expenditure history, or an insurance loss run.

The document is organized to be read front to back. Read it back to front.

### The pro forma is not your model

A broker pro forma typically achieves full market rent immediately, holds an expense ratio that the trailing statement does not support, carries the seller's tax basis forward, and exits at the same cap rate it enters at. Each of those alone is defensible as a marketing assumption. Together they compound into a net operating income that no buyer will produce.

The useful way to handle it is not to argue with it. Rebuild year one from the exhibits, then compare. The difference is a precise list of what the seller is asking you to believe, which is a far better basis for a conversation than a general sense that the numbers look aggressive.

### The exhibits are the evidence

The rent roll and the T-12 are the two documents in the package with an evidentiary character, because they report what exists rather than what could exist. A tax bill, a loss run, a utility history and a capital expenditure log are the next tier. Anything the seller declines to provide is itself information.

Reconcile the two primary exhibits against each other before you read another page of the narrative. When rent roll revenue annualized does not tie to trailing revenue, find out why.

### Rent comparables deserve real suspicion

Comparable sets are chosen. A comp set that includes three properties built fifteen years later than the subject is a comp set that produces a higher market rent. Check vintage, unit mix, amenity level, submarket and, above all, whether the quoted rents are asking rents or effective rents net of concessions.

The same applies to sale comps used to justify the cap rate. A trade from eighteen months ago in a different rate environment is a historical fact, not a pricing signal.

### What to verify outside the document

Real estate taxes at your expected reassessed basis. Insurance from an actual quote for your coverage. Market rents from your own comps or a shop. Utility costs from twelve months of bills. Deferred maintenance from an inspection rather than from the capital expenditure narrative.

Altyst reads the offering memorandum together with the rent roll, the trailing statement, the debt quote and the lease schedule, and surfaces the places they contradict each other. Marketed vacancy rarely matches the rent roll, and that disagreement is the kind of thing the software flags with the page it came from rather than quietly averaging.

Related questions

Q: What is an offering memorandum?
A: An offering memorandum is the marketing document a broker or seller prepares for a commercial property sale. It contains the property and market description, rent and sale comparables, a financial summary, and exhibits such as the rent roll and trailing twelve month operating statement.

Q: Can you trust the pro forma in an offering memorandum?
A: It should be treated as the seller's argument rather than as an underwriting. Broker pro formas commonly assume immediate achievement of market rents, an expense load below the trailing statement, the seller's existing tax basis, and an exit cap rate equal to the going-in rate.

Q: What should you read first in an offering memorandum?
A: The exhibits. The rent roll and the T-12 report what the property actually does, and reconciling them against each other before reading the narrative gives you an independent year one to compare the pro forma against.

Q: Why does marketed vacancy differ from the rent roll?
A: Marketing materials often quote physical occupancy at a favorable date or exclude units held offline, while the rent roll shows current unit-level status. Concessions, down units, model units and employee units all create legitimate differences, and each should be identified rather than averaged.

---

## What does real estate underwriting software actually do?
Source: https://altyst.ai/answers/what-is-real-estate-underwriting-software
Updated: 2026-08-06

Answer: Real estate underwriting software turns a deal's source documents into a financial model and the documents an investment committee reads. Four jobs sit inside that: extracting figures from an offering memorandum, rent roll and operating statement; computing a cash flow, debt structure and return set correctly; letting an analyst change any assumption and see everything recompute; and producing an Excel workbook and a memo that reconcile to the model. A tool that does three of the four sends you back to a spreadsheet.

### The category, stated plainly

Every acquisitions team already has an underwriting process. It runs on a spreadsheet built by somebody who left, copied per deal, and edited until it produces a number. The spreadsheet is rarely the problem. The problem is the re-keying that happens before the model starts, and the version drift that happens after it.

Software in this category is trying to remove those two problems without removing the analyst's control over the model, and those two goals pull against each other. Every template that guarantees a consistent answer also constrains the deal you can express, and every tool that lets you express anything gives up some of the consistency. Where a given product sits on that tradeoff is the thing to work out on your own deals, and it is not something a feature list will tell you.

### Where AI belongs, and where it does not

Reading a scanned rent roll is a language problem and a genuinely good use of a model. Computing an internal rate of return is not. A language model does arithmetic by predicting plausible text, which means it can produce a number that is close, confident and wrong, and it can produce a different number the second time you ask.

The line worth holding is that AI proposes values and a deterministic engine computes results. Every proposed value should arrive with the document and page it came from, and a human should confirm it before it enters the model. Determinism is checkable: run the same inputs twice and see whether you get the same answer.

### What to test before you buy anything

Take a real deal with a messy rent roll and run it. Not the vendor's demo deal.
- Does the same input produce the same output twice
- Can you edit every assumption, or only the ones the template anticipated
- Does each figure show the formula and the inputs behind it
- Does an extracted number carry its source document and page
- Does the Excel export contain live formulas, or pasted values
- Does it handle the asset classes you actually buy, not just multifamily
- Can it price the deal backwards, solving for the bid that hits your return
- What happens on a scanned PDF and on a photograph of a page

### The Excel question

Any serious tool has to export a workbook that a lender, a partner or an investment committee can open and interrogate. A PDF is a conclusion. A workbook on live formulas is an argument somebody else can check, and the ability to check it is most of why institutional real estate still runs on spreadsheets.

An export of pasted values is a screenshot with more steps.

### Where Altyst sits

Altyst reads the documents a deal already arrives with, in PDF, Excel, CSV, Word, plain text or a photograph of a page, including scanned pages through OCR. It proposes each extracted value with its source and a confidence signal for review, then a deterministic engine in exact decimal arithmetic computes the model. Editing any assumption recomputes returns, cash flow, debt and the downside case at once, and clicking a figure shows the formula behind it.

It covers multifamily, single-family and short-term rental, office, medical office, retail, industrial, flex, warehouse, self-storage, mixed-use, hotel, and land or development, each with the leasing, debt and returns that asset class needs. Outputs are an Excel workbook on live formulas, a one-page investment memo, an editable PowerPoint deck, and lender and offering packages, all reconciled to the model.

It is a paid subscription with no free tier. The current plans, the per-additional-deal rate, and each plan's monthly ceiling on new deals are all on the pricing page, which is the only place those numbers are published.

Related questions

Q: What is real estate underwriting software?
A: Software that converts a property's source documents, such as an offering memorandum, rent roll and trailing operating statement, into an editable financial model with cash flow, debt, returns and scenario analysis, and then produces the Excel and PDF outputs an investment committee reviews.

Q: Can AI underwrite a real estate deal?
A: AI is well suited to reading documents and proposing values, and poorly suited to performing the calculation. A language model produces arithmetic by predicting text, so it can return a confident wrong figure and a different figure on a second attempt. The reliable pattern is AI for extraction with a deterministic engine for the math.

Q: Does underwriting software replace Excel?
A: It replaces the re-keying and the version drift, not the workbook. Any tool used at an institutional level still has to export a workbook on live formulas, because a lender or an investment committee needs to interrogate the math rather than accept a rendered conclusion.

Q: What should you test when evaluating underwriting software?
A: Run a real deal with a messy rent roll rather than the vendor's demo. Check that identical inputs produce identical outputs, that every assumption is editable, that each figure exposes its formula, that extracted values carry a source document and page, and that the Excel export contains live formulas rather than pasted values.

---

## How much does ARGUS cost?
Source: https://altyst.ai/answers/how-much-does-argus-cost
Updated: 2026-08-06

Answer: Altus Group does not publish a price list for ARGUS Enterprise, so no official figure exists to quote: you request a quote and it is negotiated per firm. Every number that circulates on third-party directories and industry forums is second-hand, and those numbers contradict each other, from a few thousand dollars per user per year at the low end to several times that at the high end, with one-time implementation and training commonly cited in the five-figure range. Seats, modules and contract term are what drive a quote, so the only figure worth budgeting against is the one Altus puts in writing for your firm.

### Why there is no published price

ARGUS Enterprise is sold by Altus Group as part of the ARGUS Intelligence Platform, and as of 6 August 2026 there is no price on its product pages. The buying motion is a conversation: you tell them the firm, the seat count and what you model, and a quote comes back. That is ordinary for enterprise software sold into institutional real estate, and it is not a criticism. It does mean that every dollar figure you find on a search results page was written by somebody who was not the vendor.

This matters more than it sounds. When a vendor publishes a price, third-party pages either repeat it or are visibly wrong. When a vendor publishes nothing, the pages that answer the question anyway are filling a vacuum, and there is no correction mechanism. Numbers get copied from one directory to another, tiers and years get lost along the way, and a figure from an old quote for a different configuration ends up presented as the current price.

So the honest starting point is that nobody outside Altus and its customers knows what ARGUS costs, and the ranges below are evidence about the internet rather than evidence about the product.

### The figures that circulate, and why they conflict

Comparing the estimates visible on 6 August 2026, they differ from each other by roughly an order of magnitude. Both clusters below are repeated confidently, and they cannot both describe the same product on the same terms.

The most likely explanation is boring rather than sinister. Different sources are describing different things: different modules, different seat counts, an academic or single-user arrangement against an enterprise agreement, a figure from several years ago, or a total contract value divided in a way the writer did not explain. None of that is knowable from the outside, which is exactly the point.
Reported figures, second-hand and unverified:
- About $3,000 to $5,000 per user, per year | Annual license, per seat | Second-hand
- As high as roughly $1,500 per user, per month | Subscription, per seat | Second-hand
- Commonly $5,000 to $25,000, one time | Implementation and training | Second-hand
- No published figure | Altus Group's own pages | Checked 6 Aug 2026

### Treat those numbers as rumor, because that is what they are

We are reporting what circulates, not endorsing it, and we have not verified any of it with Altus. We are not quoting Altus, we are not authorized to, and nothing above should be read as a price offered by anybody. If you are building a budget, the only defensible line item is the quote you receive.

The reason this page reports the conflict at all, rather than skipping the numbers, is that the pages ranking for this question today mostly pick one of the clusters and present it as the answer. A reader who plans around the low cluster and receives a quote from the high cluster has been actively misled. Knowing that the range is wide and unverified is more useful than a confident wrong number.

### What actually drives the quote

If you are about to ask for one, these are the variables that move it, and the ones to pin down before you compare two proposals or your own renewal against last year.
- Seats. How many named users, and whether a seat is per person or floating across a team.
- Modules. What is in the SKU. ARGUS Enterprise sits inside a wider platform, so ask what is included and what is a separate line.
- Term and escalators. A one-year commitment against three, and what the price does at renewal.
- Implementation. Setup, data migration and configuration, usually a one-time fee separate from the license.
- Training and certification. Public classes, private sessions and exam fees are published separately from any license quote.
- Support tier, and whether anything you rely on sits above the standard one.

### The costs that are not on the invoice

Whatever the license comes to, two other costs decide the real total, and neither one appears on a quote.

The first is time to productivity. A modeling environment this deep is not learned in an afternoon, which is why a formal training path and a certification exist for it at all. Budget the ramp for every analyst you onboard, and budget it again when one leaves.

The second is the data entry, and it is the one nobody prices. A modeling environment starts work after the numbers are in it. Somebody still reads the offering memorandum, the broker's rent roll and the scanned trailing twelve and types all three in. That cost lands on every deal you screen, including the large majority you will pass on, and it is identical whichever modeling environment you license.

### What the alternatives cost by comparison

Three shapes of spend, side by side, so the license question sits in context.

Building the model in a spreadsheet has no license line beyond the spreadsheet program itself, and the cost is entirely labor: your template, your maintenance, and a person re-keying every document on every deal. Sending the underwriting out to an analyst or an offshore team converts it into a per-deal or per-month fee quoted per engagement, which is predictable per deal and slow when the deadline is Friday.

Altyst publishes its prices rather than quoting them: Individual $12, Professional $24, Team $99 a month, with a published price per additional deal and a published monthly ceiling on new deals for each plan. There is no implementation fee, no annual commitment and no quote to wait for, and the current numbers are always on the pricing page rather than in this paragraph.

The comparison that matters is not license against license, though. It is what one unit of spend buys. A seat license buys the environment and leaves the document reading to you. Altyst charges per deal created, and reading the documents, editing assumptions, recomputing, repricing and every export are included in that.

### How to get a number you can rely on

Ask Altus directly, and ask for the all-in annual figure rather than a per-seat rate, because a per-seat rate with implementation and training beside it is not comparable to anything.
- Ask for the all-in first-year cost: license, implementation, training, support.
- Ask what the second and third year look like, separately, including any escalator.
- Ask exactly which modules that quote covers and what a common addition costs.
- Ask what happens to your files and your access at the end of the term.
- Get every one of those answers in writing before you compare it against anything else.

### About this page

ARGUS, ARGUS Enterprise and the ARGUS Intelligence Platform are trademarks of Altus Group Limited. Altyst is not affiliated with, endorsed by, or sponsored by Altus Group, and nothing here is a quote from Altus or on its behalf. ARGUS is named because it is what people search for by name and the question deserves a straight answer.

The one verified statement on this page is that Altus published no price for ARGUS Enterprise as of 6 August 2026, which you can check yourself in under a minute. Every figure reported above is second-hand, unverified, and stated as such. Altyst sells a competing product, which is a reason to check this page against the vendor rather than to take it on trust.

Related questions

Q: Does Altus Group publish ARGUS pricing?
A: No. As of 6 August 2026 there is no price list on the ARGUS product pages. Pricing is quoted per firm, which is why every figure that appears elsewhere online is second-hand and none of it can be confirmed without asking Altus.

Q: Why do online estimates of ARGUS cost differ so much?
A: Because no published price exists to correct them. The estimates visible on 6 August 2026 differ from each other by roughly an order of magnitude, most likely because they describe different seat counts, different modules, different years, or a total contract value split in a way the writer did not explain. Treat any single figure as unverified.

Q: Is ARGUS priced per user or per company?
A: Publicly there is no answer, because Altus does not publish the structure any more than the amount. Quotes in this part of the market are usually built from seats, modules and contract term. Ask which of those your quote is priced on, and ask what happens when you add a seat mid-term.

Q: Is there an implementation fee on top of the license?
A: Enterprise deployments in this category commonly carry a one-time implementation and configuration cost separate from the license, and figures in the five-figure range are widely cited second-hand for ARGUS specifically. Neither the existence nor the size of that fee is published, so ask for it as a named line in the quote rather than assuming it is included.

Q: What does Altyst cost by comparison?
A: Altyst publishes every price rather than quoting them, on the pricing page, with a per-additional-deal rate and a monthly ceiling on new deals for each plan. There is no implementation fee and no annual commitment. It is a different unit of spend: you pay per deal created, and reading the documents, editing, recomputing and every export are included.

Q: Is ARGUS worth the cost?
A: It depends on one question, and the question is not about money. If an ARGUS file is part of what you owe a lender, an appraiser or a partner, the license is a requirement and there is nothing to evaluate. If it is not, the honest test is whether you are paying for lease-by-lease depth you actually use, or paying for it on deals you end up passing on. Screening volume and modeling depth are two different problems and they do not have to be solved by the same purchase.

# Guides

## How to underwrite a multifamily deal from an offering memorandum
Source: https://altyst.ai/learn/underwrite-a-multifamily-deal-from-an-offering-memorandum
Updated: 2026-08-06

The OM is a sales document. Underwriting it means rebuilding the income statement from the primary documents, sizing the debt three ways, and being honest about which assumption is actually carrying the return.

An offering memorandum is a sales document. That is not an accusation. A broker is paid to present the property in its best defensible light, and a good OM earns its keep: it has the rent roll, the trailing financials, the unit mix, the capital history, and a coherent story about why the asset is worth more under new ownership. What it is not is an underwriting. The pro forma at the back is the seller's answer to a question you have not asked yet.

The job is to rebuild the income statement from the primary documents, decide which of the seller's assumptions you actually believe, and then find out what the deal is worth to you, at your cost of capital, with the debt you can actually get. Here is the order that works.

## Read the package in the order the OM does not present it

An OM leads with the story and buries the evidence. Invert it. Open the rent roll first, the trailing twelve second, the capital expenditure history third, and the marketing narrative last. By the time you read the seller's rent growth assumption you should already have your own, built from the roll, and the only question left is why the two disagree.

The documents you need are almost always in the package or one email away:

- The current rent roll, unit by unit, with lease start and end dates, in-place rent, market or asking rent, concessions, and status (occupied, notice, vacant, down, model, employee).
- The T-12, ideally with the twelve monthly columns rather than only the annual total.
- A T-3 or T-6 if the property has been through a renovation or a lease-up.
- The capital expenditure history, so you know what has already been replaced and what has not.
- The current property tax bill and the insurance declarations page, not the amounts in the T-12.
- Any loan quote or assumable debt terms.

If the rent roll and the T-12 do not arrive together, ask again before you model anything. One without the other is not enough to catch the mistakes that matter.

## The rent roll is the primary document

Everything upstream of net operating income starts here. From the roll you can compute, without trusting a single number the OM asserted:

- Physical occupancy, as occupied units divided by total units. Count down units and model units honestly. A 60-unit property with two down for renovation and one model is running 57 revenue-capable units, not 60.
- Average in-place rent, weighted by unit, not the simple average of the unit-type table.
- Loss to lease, as the gap between in-place rent and the market rent the roll itself quotes. This is the single number a value-add story lives on, and it is the easiest one to inflate, because the "market rent" column is whatever the owner typed.
- Lease expiration exposure, month by month. Twelve leases expiring in the same 60 days is a real risk in a soft quarter, and it never appears in an annual pro forma.
- Concessions, which the roll often shows as a separate column and the T-12 often nets into revenue. Read both.

Take one number seriously here: verify the roll's market rent against something outside the package. If the OM says market rent is $1,675 and the comparable properties two blocks away are asking $1,575, the entire mark-to-market thesis is off by about 6%. On 60 units that is $72,000 a year of gross rent, and once you take vacancy and the management fee out of it, roughly $1.2 million of value at a 5.5% cap.

## Rebuild effective gross income line by line

Work top down, and keep every deduction visible rather than collapsing them into one "vacancy" plug. A 60-unit example, using round numbers to keep the arithmetic legible:

| Line | Amount | Basis |
| --- | --- | --- |
| Gross potential rent | $1,116,000 | 60 units at $1,550 in place, 12 months |
| Vacancy, credit loss, concessions | ($66,960) | 6.0% of GPR |
| Other income | $60,000 | $1,000 per unit per year |
| Effective gross income | $1,109,040 | |
| Operating expenses | ($559,040) | $9,317 per unit |
| Net operating income | $550,000 | |

Two things to watch in that table.

First, economic occupancy is not physical occupancy. Physical occupancy asks how many units have a body in them. Economic occupancy asks how much of gross potential rent actually reached the bank account, and it is always lower, because it absorbs concessions, bad debt, non-revenue units, and the gap between the roll's asking rent and what a signed lease actually pays. A property can be 96% physically occupied and 89% economically occupied, and only the second number pays the mortgage.

Second, other income deserves its own diligence. Utility reimbursement billing, parking, pet rent, storage, and application fees are real and recurring. A one-time legal settlement, a bulk cable rebate that has since been renegotiated, or an insurance proceeds credit is not. If other income is more than about a tenth of effective gross income, find out exactly what is in it before you grow it.

## Normalize the expenses, then add the ones the seller does not have

The T-12 shows what the property cost the current owner. It does not show what it will cost you. The recurring adjustments, roughly in order of how much money they move:

- Real estate taxes. In many jurisdictions a sale triggers a reassessment, and the trailing tax line reflects an assessed value set years ago at a lower basis. Underwrite year one taxes on your purchase price at the current millage, not on the seller's bill. In a state with a transfer-triggered reassessment this single line can move net operating income by more than every other adjustment combined.
- Insurance. The trailing premium was set at the last renewal. Get a quote. In a hardening market a premium can reprice on renewal by a margin that swamps your entire rent growth assumption.
- Management fee. Almost always a percentage of effective gross income. An owner who self-manages may show zero or a token amount. Underwrite the fee in your actual management agreement, and underwrite it on your effective gross income, not the seller's.
- Payroll. On a portfolio, on-site staff are frequently allocated across several assets. Ask how many full-time equivalents actually sit at this property.
- Replacement reserves. Almost never in a seller's T-12, always required by a lender, and always required by reality. Reserve per unit per year is a negotiation with your lender, not a constant, but underwriting zero is underwriting a fiction.

Then compare the normalized expense load two ways: as a ratio to effective gross income, and as dollars per unit per year. The ratio catches structural problems. The per-unit figure catches a property whose expenses look fine only because its rents are high. If either lands far outside the range for the submarket and vintage, you have either found something or missed something, and both are worth an hour.

Altyst reads the rent roll and the T-12 and ties the two against each other (/product/documents). Where the stated net operating income will not reconcile with the rent and expense lines it is supposedly built from, the figures involved come back with lowered confidence and a note about what does not add up, which is where the unit-of-measure mistakes hide.

## Your basis is not the asking price

Going-in cap rate is net operating income divided by total basis, not by the purchase price. Basis includes capitalized closing costs, the acquisition fee if there is one, and any day-one capital you have to spend before the property performs. On a $10,000,000 purchase with 1.5% in closing costs, $550,000 of net operating income is a 5.50% cap on price and a 5.42% cap on basis. That 8 basis points is small. On a deal with a $1,200,000 renovation budget it is not small at all, and quoting the cap on price is how a sponsor accidentally overstates yield to an investment committee.

The same applies to yield on cost in a value-add: the denominator is everything you will have put in by stabilization, including the capital you spend in year two.

## Let the debt size itself

Never start from the loan amount you want. Size the loan three ways and take the smallest, because that is what a lender will do:

1. Loan to value. At 65% of a $10,000,000 value, $6,500,000.
2. Debt service coverage. At a 6.0% rate on a 30-year amortization schedule, the annual constant is about 7.19%. A 1.25x minimum coverage on $550,000 of net operating income allows $440,000 of annual debt service, which supports about $6,115,000.
3. Debt yield. At a 9.0% floor, $550,000 of net operating income supports about $6,111,000.

Debt yield binds, at roughly $6,111,000. Note what happened: the LTV test, the one most sponsors quote, was the loosest of the three and never mattered. Note also the second thing those three numbers are telling you. The loan constant is 7.19% and the going-in cap rate is 5.50%. That is negative leverage: the debt costs more than the asset yields on day one, so borrowing more lowers your going-in cash return rather than raising it. Negative leverage is not automatically disqualifying, and in a low-cap market it is common, but it means every dollar of return is coming from growth and exit rather than from current income, and you should say so out loud rather than let a levered internal rate of return hide it.

While you are here, compute break-even occupancy: operating expenses plus debt service, divided by gross potential rent plus other income. In this example that is $559,040 of expenses plus about $440,000 of debt service, over $1,176,000 of potential revenue, or roughly 85%. That is your margin of safety expressed in a form a lender and a partner both understand instantly.

## The exit assumption is doing most of the work

In a five-year hold on a levered deal, the reversion is usually well over half of the total return. Which means the exit cap rate, a number nobody can know, is the single largest driver of an answer you are about to present as precise.

Two disciplines help.

Capitalize forward net operating income. The buyer in year five is pricing year six income, not year five income. Growing $550,000 at 3% for five years gives $619,030 in year five and $637,601 in year six. At a 5.75% exit cap that is a value of about $11,089,000, and after a 2% cost of sale, roughly $10,867,000 of gross proceeds. On a $10,150,000 basis, that is a thin gain, and the deal has to earn its return from cash flow and from executing the mark-to-market, not from the sale.

Expand the exit cap above the going-in cap. The property is five years older and the market may not be. Holding the exit equal to the entry cap is an assumption that the world stays exactly where it is, which is a forecast, not a base case. A common house rule is to set the exit at least 25 basis points wider than going in, and it is a good default even when nobody is making you do it.

## Stress it before you believe it

A single-point answer is not an underwriting. Before the deal goes anywhere, run at least these:

- Exit cap plus 50 and plus 100 basis points.
- Rent growth at zero for the first two years.
- The renovation premium at half of what the seller achieved.
- Interest rate up 100 basis points at refinance or on the floating tranche.
- A two-way sensitivity grid on exit cap against rent growth, because those two are correlated in the real world and a one-at-a-time sweep will understate the downside.

What you are looking for is the point where the deal stops working, and how far away it is from your base case. Get a feel for the units first. At a 5.75% exit cap, a quarter point of expansion takes about 4% off the exit value, and at 65% leverage that 4% is roughly 12% of your equity. If the deal only clears your hurdle while the exit cap sits exactly where the entry cap sits, it is a bet on cap rates wearing a value-add costume.

## The order of operations, compressed

1. Rent roll: occupancy, weighted in-place rent, loss to lease, expiration schedule.
2. T-12: normalize to a forward-looking expense run rate.
3. Reset taxes to your basis, insurance to a live quote, management to your contract, and add reserves.
4. Build effective gross income and net operating income with every deduction visible.
5. Compute the going-in cap on total basis, not on price.
6. Size debt on LTV, coverage, and debt yield, and take the binding constraint.
7. Check for negative leverage and compute break-even occupancy.
8. Set the exit on forward net operating income at a cap wider than going in.
9. Stress the exit, the growth, and the rate, together and not one at a time.
10. Write down, in one sentence, which assumption the return depends on most.

That last step is the one people skip, and it is the one an investment committee will ask about first. If you cannot name the assumption carrying the deal, you have not underwritten it yet. You have recalculated the OM.

> Altyst runs this whole sequence from the documents you already have. Extraction proposes the values with their source and confidence, a deterministic engine does every calculation (/product/modeling), and the model recomputes end to end the moment you change an assumption. See it on a sample deal (/demo).

---

## What a T-12 hides, and how to normalize it
Source: https://altyst.ai/learn/what-a-t-12-hides-and-how-to-normalize-it
Updated: 2026-08-06

A trailing twelve is a record of what the property cost somebody else, under their tax basis, their insurance policy, their management contract, and their capitalization habits. None of those survive the closing.

The trailing twelve gets more trust than any other document in a deal package, and it deserves some of it. It is actual money that actually moved. But a T-12 answers a narrow question: what did this property cost the current owner, under their assessed value, their insurance policy, their management agreement, their payroll allocation, and their opinion about what counts as capital. You are not buying any of those. You are buying the building.

Normalization is the work of turning that record into a forward-looking run rate. Here is what hides in the document, roughly ordered by how much money it moves.

## Read the twelve months, not the total

Ask for the T-12 with monthly columns. The annual total conceals almost everything interesting.

Monthly columns show you seasonality (utilities, turnover, snow removal), they show you the month a line item stepped up and stayed there, and they show you the one month with a $40,000 repair sitting in the middle of an otherwise flat maintenance line. They also let you compute a T-3 and a T-6 annualized, which is the honest way to read a property that has just come through a renovation or a lease-up.

Trend statements cut both ways, and knowing which way is the skill:

- A T-3 annualized is the right revenue basis for a property whose rents have genuinely stepped up, because the trailing twelve is diluted by nine months of pre-renovation rent.
- A T-3 annualized is the wrong expense basis for anything seasonal. Three summer months annualized will understate heating and overstate cooling, and three months that happen to miss the annual insurance and tax accrual will understate both to zero.

The usual practice, and a defensible one, is to take revenue on the most recent trend and expenses on the full twelve, then adjust the expense lines individually. Say which you did, and why.

## Property taxes are the biggest single miss

In many jurisdictions, a sale resets the assessed value. The trailing tax line reflects an assessment made under the prior owner, often years ago, at a lower basis, sometimes with an abatement or an exemption that does not transfer.

Underwrite year one taxes as your purchase price times the current effective millage, then check whether the jurisdiction reassesses on transfer, on a cycle, or on appeal. In a transfer-triggered state this one adjustment routinely moves net operating income more than every other normalization combined, and it moves it in the direction that costs you money.

The related trap is the abatement schedule. A property in year three of a ten-year phased abatement has a tax line that is scheduled to grow every single year, regardless of assessment. That growth belongs in the model as a specific per-year figure, not as a general expense growth rate.

## Insurance is priced at renewal, not at closing

The premium in the T-12 was set at the last renewal, under the prior owner's loss history, deductible, and portfolio placement. Yours will be different, and in a hard property market the difference can be a multiple rather than a few percent.

Get a live quote before the model is final. If you cannot, underwrite the line at a level you would be comfortable defending after a bad renewal, and flag it as an assumption rather than a fact. Insurance is one of the few operating lines that can reprice by a large percentage in a single year, which makes it far more dangerous than its share of the expense load suggests.

## Management fees, payroll, and the owner who works for free

Three related distortions:

- Management fee. Usually a percentage of effective gross income. A self-managing owner may show a token fee or none. Replace it with the fee in your actual agreement, computed on your effective gross income.
- Payroll. On-site staff are often shared across a portfolio and allocated by a formula that has nothing to do with the hours actually worked at this property. Ask for the headcount that physically sits here.
- Owner labor. A local owner who does turns himself shows a repairs line that no third-party manager can reproduce. That is not a savings you inherit.

The same logic applies in reverse. An owner whose brother-in-law does landscaping at above-market rates leaves you an expense line that really can come down, and that is a legitimate underwriting adjustment, so long as you can name the contract.

## Repairs that are really capital, and capital that is really repairs

This is where two properties running identically can report expense loads that are nowhere near each other, entirely on accounting policy.

Owners differ on where the line sits between a repair and a capital improvement. One capitalizes every unit turn above a threshold; another expenses all of it. One capitalizes the roof; another expenses "roof repairs" that added up to the same roof. Neither is lying. But if you take a T-12 from an aggressive capitalizer and grow it, you are underwriting a maintenance budget that has been quietly pushed into a capital account you are not funding.

The fix is mechanical:

1. Get the capital expenditure schedule alongside the T-12.
2. Pull anything in the operating statement that is clearly capital in nature and move it out.
3. Push anything in the capital schedule that is really recurring maintenance back into operating expenses.
4. Then, separately, fund a replacement reserve, because the reserve is a claim on future cash regardless of how the last owner classified anything.

Step four is the one people argue about. Whether reserves sit above or below net operating income is a convention that varies by lender and by shop. What is not optional is funding them somewhere. A model with no reserve line is a model that assumes the roof is eternal.

## One-time items hide in plain sight

Scan the twelve monthly columns for any line that is flat for eleven months and spikes in one. Common finds:

- A legal settlement or an eviction wave.
- A storm deductible, or insurance proceeds booked as a revenue credit.
- A prior-period adjustment, which is an accounting correction for a mistake made in a year you are not buying.
- A one-time utility rebate or a bulk-service signing bonus.
- A tax refund from a successful appeal, which is real money but does not repeat.

Each one gets removed from the run rate. Each one also gets noted, because a pattern of them tells you something about the asset.

## Revenue is billed, collected, and forgiven, and only one of those is real

Operating statements are inconsistent about where they show revenue deductions. Some present gross potential rent and then subtract vacancy, loss to lease, concessions, and bad debt as visible lines. Others present a single net "rental income" figure with all of that already inside it.

You need the gross-to-net bridge to underwrite anything, because the seller's growth story lives in it. If concessions are one month free on a twelve-month lease, that is roughly 8% off effective rent, and whether it appears as a concession line or is simply absorbed into a lower net rental income figure changes the apparent loss to lease dramatically.

Insist on the components. Then compute economic occupancy as collected rental revenue over gross potential rent, and compare it to the physical occupancy on the rent roll. A wide gap is a story: heavy concessions, chronic bad debt, or a rent roll whose "market rent" column is aspirational.

## Below the line is not your business

A seller-prepared statement often includes items that do not belong in net operating income at all: debt service, depreciation and amortization, asset management fees paid to the sponsor, partnership administrative costs, owner draws, and capital reserve contributions.

Strip them. Net operating income is a property-level measure, deliberately independent of how the asset is financed and who owns it, which is exactly what makes it comparable across deals. Leaving an asset management fee inside operating expenses does not make you conservative; it makes your cap rate incomparable to every other cap rate you are looking at.

Then add the property-level costs the seller does not have and you will: your reserve, your management fee, and any compliance or reporting cost specific to your structure.

## Build a normalization column, not a normalized number

The output of this work should never be a single adjusted expense figure. It should be a table with four columns: the trailing amount, the adjustment, the normalized amount, and a one-line reason.

The reason to bother is that a partner can then argue with a specific row instead of with your judgment in general, which makes for a much shorter meeting. It also survives diligence: when the tax certiorari consultant comes back with a different assessment, you change one line rather than rebuilding the sheet. And a year later you can still read what you assumed, which is the only way anybody gets better at this.

This is the same discipline the model itself should enforce. Every extracted figure in Altyst carries the document it was read from and a confidence signal (/product/auditability), and every output shows the formula and the inputs behind it, so a normalization adjustment stays visibly an adjustment rather than a number somebody typed into a spreadsheet in 2023.

## Tie it back to the rent roll before you trust it

The final check, and the one that catches the most errors: take your normalized annual rental revenue, divide by twelve, divide by the number of occupied units on the rent roll, and compare the result to the weighted average in-place rent from the roll itself.

Those two numbers come from completely different documents produced by different systems. If they agree within a couple of percent, your normalization is probably sound. If they do not, one of four things is true: you have the wrong occupancy, the roll includes non-revenue units, the T-12 revenue line has something in it that is not rent, or somebody made a unit-count error. All four are worth finding before you send a letter of intent, and none of them will announce themselves.

> Altyst runs this reconciliation on the documents themselves. The rent roll, the trailing statement, and the offering memorandum are extracted and then tied out against each other, and where they will not reconcile the figures involved come back with lowered confidence and a note about what does not add up, rather than one reading quietly winning. See how document intelligence works (/product/documents), or read the full calculation methodology (/resources/methodology).

---

## The after-tax return math most underwriting skips
Source: https://altyst.ai/learn/after-tax-returns-depreciation-recapture-and-the-1031
Updated: 2026-08-06

Almost every institutional model quotes a pre-tax internal rate of return. That is the right convention and the wrong number for the person actually writing the check.

Almost every institutional real estate model quotes returns pre-tax. There is a good reason for it. A fund with limited partners in different states, different entities, and different personal tax positions cannot compute one after-tax number that is true for all of them, so the property-level pre-tax return is the only figure everyone can compare. Pre-tax is the right convention for a syndicated deal.

It is also the wrong number for a decision, and the gap between the two is not small. For an individual or a family office buying in their own name or through a pass-through, the after-tax return can land well below the headline. The worked example at the end of this piece gives up about 250 basis points of internal rate of return, and how wide that gap gets depends on asset class, leverage, hold period, and how much of the basis is land. Two deals with the same pre-tax internal rate of return can be quite different investments after tax.

Here is the mechanism, piece by piece.

## Depreciation is the whole reason real estate shelters income

The depreciable basis is the purchase price plus capitalized closing costs, less the value allocated to land, plus every capital dollar you spend afterward. Land is never depreciated, because it does not wear out. Improvements are, on a straight line, over a statutory recovery period: 27.5 years for residential rental property and 39 years for nonresidential real property.

The land allocation matters more than people expect. A deal where land is 20% of basis depreciates 80 cents of every dollar; at a 35% land allocation it depreciates 65 cents, which is about 19% less deduction in every year of the hold on the same purchase price. That difference flows straight through to taxable income. It is also one of the few tax variables you can influence at closing, through the purchase price allocation and an appraisal that supports it.

Taxable income from operations is then:

> net operating income, less interest, less depreciation

Two things in that line surprise people the first time. Principal is not deductible, only interest is, so a fully amortizing loan produces taxable income that grows faster than cash flow as the amortization curve shifts. And depreciation is not cash, so a property can distribute money every quarter while reporting a tax loss. That is the shelter, and it is the reason levered real estate is a tax-efficient asset in the first place.

## Cost segregation and bonus depreciation move the deduction forward

A cost segregation study reclassifies part of the improved basis out of the 27.5 or 39 year bucket and into shorter-life personal property and land improvements, typically 5, 7, and 15 year property. Under bonus depreciation rules, the eligible slice can be expensed heavily or entirely in the year the property is placed in service rather than spread over decades.

The applicable bonus percentage depends on the placed-in-service year and on current law, which has changed repeatedly. Do not model a rate from memory. Use the rate your accountant gives you for your tax year, and use the eligible fraction your study actually supports rather than a rule of thumb.

There is a real cost to the acceleration, and it is the part most cost segregation pitches move past quickly. The 5 and 7 year personal property is section 1245 property, and when you sell, its recapture is taxed at ordinary income rates, not at the 25% ceiling that applies to straight-line depreciation on real property. The 15 year land improvements are section 1250 property rather than 1245, but they are written off on a 150% declining balance schedule, so the depreciation taken above straight line comes back as ordinary income as well. Either way, most of what you accelerated returns at your ordinary rate. A cost segregation study converts a deduction taken at your ordinary rate today into a recapture at your ordinary rate later. The benefit is time value and nothing else.

Time value usually wins, especially on a long hold or when the deduction offsets income taxed at a high marginal rate. But it does not always win, and a model should show you the crossover rather than assume it. If you plan to sell in three years, the deferral window is short and the recapture arrives fast.

## A loss you cannot use is not lost

When depreciation drives taxable income negative, a passive investor generally cannot deduct that loss against wage or portfolio income. It is suspended under the passive activity loss rules and carried forward.

Suspended losses are not wasted. They do two things:

1. They offset future positive taxable income from the same activity as the shelter thins out later in the hold.
2. They release on a fully taxable disposition of the activity, offsetting the gain when you sell. A 1031 exchange is not one, so an exchange carries the suspended losses forward alongside the deferred tax.

This is why a five-year cash flow table showing zero tax in years one through three is usually correct rather than a modeling error, and why the tax bill lands almost entirely at the sale.

The exceptions matter and are individual. Real estate professional status, material participation, the short-term rental treatment, and the special allowance for an actively participating small landlord, which phases out as income rises, all change the answer. None of that is knowable from the property, which is precisely why an after-tax model should take rates and treatment as inputs rather than assert them.

## The bill at sale has three parts

At disposition, the taxable gain is the net sale price less the adjusted basis, where adjusted basis is your original basis reduced by all the depreciation you took. That gain then splits:

- Unrecaptured section 1250 gain. The portion of the gain attributable to straight-line depreciation on the real property, taxed at a federal rate capped at 25%.
- Section 1245 recapture. The portion attributable to the short-life property from a cost segregation study, taxed at ordinary rates.
- Capital gain. Everything above that, which is genuine appreciation, taxed at long-term capital gains rates.

Layered on top, depending on your situation: the net investment income tax, and state income tax, both of which can be folded into the rates you use rather than modeled separately.

The point worth internalizing is that depreciation is a loan, not a gift, unless you never sell. You deduct at your ordinary rate during the hold and pay it back at up to 25% on the straight-line portion, which is still a good trade for most taxpayers, but it is a trade rather than free money.

## The 1031 exchange defers, it does not forgive

A properly executed like-kind exchange defers both the recapture and the capital gain by carrying your old basis into the replacement property. The tax is not eliminated. It is moved.

Two consequences follow, and both are routinely left out of models:

- The replacement property inherits a low carryover basis, which means less depreciation on the new asset for the rest of its life. You have traded a tax bill today for a smaller shelter tomorrow.
- The deferred liability is still yours. It reappears on a future taxable sale, and it compounds across a chain of exchanges.

An honest model reports the deferred amount rather than quietly deleting it. Deferral is worth a great deal, sometimes an enormous amount, but what it is worth is the time value of the deferred tax, not the tax itself.

## A worked example

Round numbers, a residential asset, one investor, interest-only debt, held five years. This is an illustration built to be arithmetically consistent, not a market forecast, and the rates are placeholders for whatever your own situation produces. It also takes a full year of depreciation in year one instead of applying the mid-month convention, which a real return would not.

Setup

| Input | Value |
| --- | --- |
| Purchase price | $10,000,000 |
| Land fraction | 20% |
| Depreciable basis | $8,000,000 |
| Recovery period | 27.5 years, residential |
| Annual depreciation | $290,909 |
| Loan, interest only | $6,500,000 at 6.0% |
| Annual interest | $390,000 |
| Equity | $3,500,000 |
| Year 1 net operating income | $550,000 |
| Net operating income growth | 3.0% a year |
| Sale price, end of year 5 | $12,000,000 |
| Cost of sale | 2.0% |

Operations

| Year | NOI | Interest | Levered cash flow | Depreciation | Taxable income |
| --- | --- | --- | --- | --- | --- |
| 1 | $550,000 | $390,000 | $160,000 | $290,909 | ($130,909) |
| 2 | $566,500 | $390,000 | $176,500 | $290,909 | ($114,409) |
| 3 | $583,495 | $390,000 | $193,495 | $290,909 | ($97,414) |
| 4 | $601,000 | $390,000 | $211,000 | $290,909 | ($79,909) |
| 5 | $619,030 | $390,000 | $229,030 | $290,909 | ($61,879) |

Taxable income is negative in every year, so no operating tax is due, and $484,520 of suspended passive losses accumulate. The investor collects $970,025 of cash over five years and reports a cumulative tax loss. That is the shelter working exactly as intended.

Sale

Accumulated depreciation over five years is $1,454,545, so the adjusted basis is $8,545,455. Net sale price after a 2% cost of sale is $11,760,000. Total gain is $3,214,545.

The suspended losses release against that gain, leaving $2,730,025 taxable. At a 25% recapture rate on the $1,454,545 of accumulated depreciation and a 20% capital gains rate on the $1,275,480 balance, tax at sale is $618,732.

Result

| Measure | Pre-tax | After-tax |
| --- | --- | --- |
| Internal rate of return | about 13.2% | about 10.7% |
| Equity multiple | 1.78x | 1.60x |
| Total tax | | $618,732 |

A 250 basis point drag, all of it at the sale, on a deal that paid no tax at all for five years.

Note what the example leaves out: the net investment income tax and state income tax, either of which widens the drag. Note also a convention it does apply. The released suspended losses are netted against the gain before the gain is split into its recapture and appreciation slices, which is how the model treats them, and it means the losses land against the 20% slice here. A released passive loss is an ordinary deduction, so a taxpayer with other income to absorb it may do better than this table shows. Each of those moves the number, which is the argument for computing it against your own facts rather than estimating it.

## What after-tax math changes about a decision

Once you can see both numbers, several comparisons stop being close calls.

- Asset class. A residential asset depreciating over 27.5 years shelters more per dollar of basis than a commercial asset over 39. Two deals at the same pre-tax return are not the same deal.
- Land allocation. A high-land-value urban site generates less depreciation than a suburban asset at the same price. This is a real and permanent difference in after-tax yield.
- Leverage. Interest is deductible, and depreciation does not shrink because you borrowed, so the same dollars of shelter sit on a smaller slice of equity. Leverage does more for the after-tax return than the pre-tax comparison suggests.
- Hold period. A longer hold accumulates more depreciation and therefore more recapture, but it also defers that recapture further into the future. The two effects run in opposite directions and the crossover is deal-specific.
- Exit strategy. Whether you plan to exchange, sell, or hold indefinitely changes the answer more than most operating assumptions do, and it is usually decided last.

## Where a model has to admit what it does not know

After-tax analysis is only honest when it is explicit about its inputs. Your marginal rate, your entity, your basis, your participation status, and your state are not properties of the building. A model that hardcodes a tax rate is producing a number that is wrong for almost everybody.

That is the design Altyst uses. The pre-tax underwriting stays the headline, because it is the comparable figure and the institutional standard. The after-tax view is a per-deal overlay that is off until you turn it on, and it uses the rates you enter. The depreciation schedule, the after-tax internal rate of return, the total tax over the hold, and the tax due at sale each open into the formula and the terms behind them. The year by year table shows net operating income, interest, depreciation, taxable income and tax for every year of the hold. And if you elect a 1031, the deferred amount is stated on screen rather than deleted.

Altyst produces model outputs for screening and analysis. It is not tax advice, and no model replaces your accountant. What a model can do is put the after-tax number next to the pre-tax number so you know how big the gap is before you sign, instead of after.

> Read the full calculation methodology (/resources/methodology), see how the modeling engine works (/product/modeling), or open the sample deal (/demo).

# Comparisons

These pages compare ways of doing the work, so a comparison lifted from here is a comparison
of approaches. Two products are named on this site and only two, the spreadsheet and ARGUS,
in both cases by nominative use with a non-affiliation notice and no logo; every statement
about ARGUS is either sourced to the vendor's own published pages or is an instruction to go
ask the vendor, and the ARGUS pages tell the reader when to keep it. No rival's price is
stated as fact anywhere on this site: the one page that answers what a competing product
costs reports that its vendor publishes no price list and that the figures circulating
online are second-hand and contradict each other. Each one
states where Altyst is the wrong answer as plainly as where it is the right one.

## Altyst as an ARGUS alternative
Source: https://altyst.ai/vs/argus
Updated: 2026-08-06

ARGUS is a modeling environment. Altyst is what turns the documents into a model. If you are here because the renewal quote landed, read the first section before you read anything else, because there are deals where the answer is that you should keep it.

### Stay on ARGUS when

- The deliverable is an ARGUS file. If an appraiser, a lender, a JV partner or an acquiring fund has asked for one by name, nothing else answers that request, and this page will not pretend otherwise.
- You are modeling institutional office, retail or industrial lease by lease, at the level of detail where every option, recovery structure and reimbursement method has to be expressed exactly as the lease reads.
- Your team is already trained. Altus publishes a training path and a certification exam, and the practical effect is real: a new analyst who holds one arrives able to open your model on day one.
- The house valuation convention is encoded in ARGUS assumptions your investment committee reads on sight. Nothing imports institutional memory.

### Bring in Altyst when

- The bottleneck is the first two hours, not the model. A PDF offering memorandum, a broker's own Excel rent roll and a scanned T-12 arrive, and somebody is retyping all three before any modeling starts.
- You screen far more deals than you bid on, and most of the work goes into proving that a deal is a pass.
- You underwrite across asset classes, including the ones a lease-by-lease model was never built for: multifamily, single-family, self-storage, hotel, and land or development.
- Six weeks later, at the investment committee, somebody asks where the insurance line came from and you want to answer from the document rather than from memory.
- You still need the Excel workbook at the end. You get one, and its core tabs recalculate.

## Start with what ARGUS is genuinely good at

Anyone selling you an alternative will skip this part. It is the most important section on the page, because if you do not know what you would be giving up you cannot tell whether the trade is worth making.

ARGUS Enterprise, sold by Altus Group as part of the ARGUS Intelligence Platform, has been the vocabulary of institutional commercial valuation for a very long time. That word, vocabulary, is the point. When an appraiser, a lender and a buyer all model the same office building, they are not only running the same arithmetic. They are using the same names for the same things: the recovery structure, the market leasing assumption, the rollover profile, the option nobody has exercised yet. Agreement on vocabulary is what lets three parties argue about one number instead of about four models.

Which produces the fact that settles this whole question for some readers. On institutional office, retail and industrial deals the ARGUS file is frequently part of the deliverable. If somebody has asked you for one by name, there is no alternative to evaluate. Renew, and stop reading here. This page is not going to talk you out of a contractual requirement, and any page that tries is not being straight with you.

The depth is real too. Lease by lease, with options, recoveries, reimbursement methods and expense stops expressed the way the lease actually reads, is genuinely hard to model, and that product has had decades to get it right.

And the training is an asset rather than an overhead. Altus publishes a two-day public class aimed at new-to-intermediate users, private training, self-paced e-learning, and a certification exam it describes as an industry recognized designation. A firm that has put a team through that has bought interchangeability: any certified analyst can open any model. That is a real return, and it is why a renewal is often the rational call even at a firm that grumbles about the invoice.

Hold on to all of that. The case for anything else has to be made against it, not around it.

## Where the friction actually is

Notice what none of the above is about. Not one of those strengths touches the first two hours of a deal.

A package arrives. A PDF offering memorandum on the broker's layout. A rent roll in Excel, in whatever columns that broker happens to use. A trailing twelve that was printed, signed, and scanned back in at a slight angle. A debt quote in the body of an email. Every figure that reaches your model gets there because a person read it off one document and typed it into another.

That is true whichever modeling environment you use. No modeling environment solves it, because the job of a modeling environment starts after the data is in. The model is empty until somebody fills it.

So the two costs are these. First, the retyping, which is where transposition errors happen, and they happen on exactly the numbers that matter, because the numbers that matter are the ones there are most of. Second, the screen. You look at far more deals than you buy, and most of the underwriting work in any acquisitions shop goes into establishing that a deal is a pass. A seat licensed and trained for the deals you win is an expensive way to prove that thirty deals were not worth winning.

## What Altyst does instead

Upload the offering memorandum, the rent roll and the T-12. A full underwriting model comes back, and every number in it is editable.

- It reads PDF, Excel, CSV, Word, plain text, and a photograph of a page. Scanned documents go through optical character recognition, because the T-12 always turns out to be a scan.
- Extracted values arrive as proposals, not as facts. Each one carries its source and a confidence, and you accept or overwrite it before it enters the model. The documents are also tied out against each other, so when the rent roll and the trailing twelve disagree about in-place income the conflict comes to you rather than being quietly resolved by whichever document was parsed last.
- Every financial result is computed by a deterministic engine in exact decimal arithmetic. Not by a language model predicting a plausible number. The same inputs always produce the same result, which is the only property that makes any downstream audit possible.
- Every extracted figure traces back to the document it came from and the excerpt it was read from, with a page number when that excerpt can be placed on exactly one page. Click a number and see the formula underneath it.
- Change any assumption and returns, cash flow, debt and the downside case recompute together, in seconds, as many times as you like. Repricing is never metered.

The sentence that matters: nobody types the rent roll in.

Underneath, the model is not a thin one. Debt sizes itself on a loan-to-value cap, on a minimum debt service coverage test, or on a debt yield floor, with mezzanine, interest-only periods, additional tranches, and a bridge to refinance. Rollover carries TI, LC, downtime, renewal probability and expense stops. The partnership waterfall does a preferred return, return of capital, an optional catch-up, a residual promote and multiple internal rate of return hurdles. There is an after-tax overlay with straight-line and bonus depreciation, passive-loss carryforward, and depreciation recapture plus capital gains at sale, with an optional 1031 exchange, at rates you supply. The waterfall and the after-tax overlay are both off until you turn them on, because a model that silently applies a promote or a tax rate you did not choose is worse than no model at all.

The asset classes run on four adapters built for four different income models rather than on one model with the labels swapped: per-unit residential for multifamily, single-family and short-term rental; lease-by-lease commercial for office, medical office, retail, industrial, flex, warehouse, self-storage and mixed-use; development carry for land, redevelopment and adaptive reuse; and a departmental hotel model on occupied room nights and ADR. Inside the commercial adapter each class carries its own recovery, vacancy, exit and lease economics, so self-storage books no TI and no leasing commissions and recovers nothing, where office recovers most of its operating expenses and carries months of downtime on rollover.

## Side by side, on the things that decide it

| The job | ARGUS Enterprise | Altyst |
| --- | --- | --- |
| Getting the figures in | Ask what is included. Altus advertises import and mapping for structured files under the Intelligence Platform, so what you get depends on the SKU you would be quoted | Upload the documents. Values are proposed with their source document, the excerpt they were read from and a confidence, for you to accept or overwrite |
| Setup time | License per seat, provision, then the model gets built from entered data | Sign up, upload a package, the model comes back. Nothing to install |
| Learning curve | Altus publishes a full path: a two-day public class for new-to-intermediate users, private training, self-paced e-learning, and a certification exam | The first model comes back before you have learned anything. The depth is there when you go looking for it |
| Price | Not published. You request a quote from Altus. What that costs, and why every figure you find online is second-hand, is covered in what ARGUS costs (/answers/how-much-does-argus-cost) | Published on the pricing page (/pricing): Individual $12, Professional $24, Team $99 a month. Every plan states its monthly ceiling on new deals. No annual commitment, no implementation fee |
| What one unit of spend buys | A license, on whatever terms the quote sets | One deal creation. Edits, recomputes, repricing and every export are never metered, and unused deal credits roll over |
| Excel export | Reports export to Excel | A workbook whose core tabs carry live formulas: EGI, NOI, levered cash flow, and an IRR over the cash-flow range. The Cover tab names which tabs recalculate and says plainly that the rest are pasted values |
| Other outputs | Reporting and portfolio tooling, varies by SKU | Six export surfaces: Excel workbook, memo PDF, editable PowerPoint deck, deck PDF, lender package, offering summary |
| Audit trail | The model records what was entered | On screen, every extracted figure carries its document, the excerpt it was read from and a confidence, plus a page when the excerpt can be placed on one. In the workbook, a Provenance tab listing every driving assumption with the kind of source behind it |
| Collaboration | Depends on the SKU. ARGUS Enterprise sits inside the ARGUS Intelligence Platform, so ask specifically what your quote includes | A shared workspace, with the seat count set by the plan. Everyone sees the same deals. No real-time co-editing of a single cell |
| Asset classes | Deepest on lease-by-lease commercial | Multifamily through land and development, on four adapters built for four different income models rather than one model with the labels swapped |
| Best for | Institutional lease-by-lease valuation, and any deal where an ARGUS file is the deliverable | Turning a document package into a defensible model fast, across asset classes, at screening volume |

Two honest notes on that table. The ARGUS column is deliberately thin on anything that cannot be checked against Altus's own published pages, because a comparison table that guesses at a rival's current feature set is wrong within a quarter and wrong in our name. And the price row carries no number for ARGUS because Altus publishes none. The figures that circulate on third-party sites differ from each other by roughly an order of magnitude, which tells you exactly what they are worth. Get the quote.

## The three questions a skeptical analyst asks

### Can I trust the numbers?

The right thing to be suspicious of is the phrase "AI underwriting", and you should stay suspicious of it, including here. So, precisely.

AI does one job on this platform: it reads documents and proposes values. It never performs the calculation. Every financial result comes from a deterministic engine doing exact decimal arithmetic, so the same inputs produce the same output on every run. Run the same deal twice on any product you are evaluating and compare. If a returned IRR moves between two runs on identical inputs, nothing downstream of it can be audited, not the memo, not the committee deck, not the answer you give a lender.

Extraction is the part that can be wrong, which is exactly why an extracted value is a proposal rather than a fact. It shows you what it read, where it read it, and how confident it was, and then it waits for you. Documents are tied out against each other and disagreements come to you. Every model carries a Model Checks tab, and a separate evaluator in our test suite re-reads the exported workbook's own formulas, computes them, and fails the build if any of them disagrees with the engine.

What we do not have: SOC 2, ISO 27001, or a comparable certification. We say so here and on the security page (/security), and you should ask every vendor on your list the same question and get the answer in writing.

The only test that settles this is yours. Run a deal you have already underwritten by hand and compare it line by line.

### Can I export the model to Excel?

Yes, and this is the part most people do not expect from a product in this category.

The exported workbook carries live formulas on its core tabs. EGI is gross potential rent plus other income less vacancy and credit loss. NOI is EGI less operating expenses. Levered cash flow is unlevered cash flow less debt service. The IRR row is a real IRR over the cash-flow range. Open it in Excel, change a cell on a live tab, and the sheet recalculates the way a model you built yourself would.

Every workbook carries Cover, Model, Cash Flow, Debt, Scenarios, Sensitivities, Provenance and Model Checks, and a deal picks up more depending on what it needs, among them Sources and Uses, Reprice, Partnership, After-Tax, Rent Roll, Tenant Schedule, Risks and Diligence. The Provenance tab works at the level of the assumption rather than the cell: every driving assumption, its value, whether it was your figure or one the system proposed, what kind of source stands behind it, and how firm that source is. The document-and-excerpt trail for a specific extracted figure is on screen.

Two things worth knowing, and worth asking every other vendor:

1. Not every tab recalculates. Some are computed by the engine and written in as values. The Cover tab names which tabs carry live formulas and states plainly that the rest do not, so nobody discovers it by editing a cell and watching nothing move. That line is generated by scanning the finished file for formula cells rather than typed by hand, so it cannot claim a tab is live when its numbers are pasted.
2. The formulas are checked, not assumed. The evaluator described above re-reads them and reconciles them to the engine on every test run.

Ask everyone on your shortlist for a sample export before you sign anything, and open it. Whether the cells contain formulas or numbers is the fastest single test of how seriously a product takes handing the work back to you.

### What about my existing models?

Keep them, and hear the limitation plainly: Altyst does not import ARGUS files, and it does not import your firm's Excel template. You should not find that out on day three.

What that means in practice:

- If an ARGUS file is a deliverable you owe somebody, that requirement does not move. Keep the license for the deals that carry it.
- The common pattern is not replacement, it is sequencing. Screen and underwrite in Altyst, where the documents do the data entry. Export the workbook. Carry the handful of deals that survive the screen into whatever your firm uses for the deals it actually bids on.
- Your house convention lives in your template, and no import wizard has ever moved institutional memory. That is an argument for keeping the template, not for retyping every rent roll into it.
- Nothing is trapped. The exports are files. They open without us, and they stay accessible if you cancel. Only new paid processing pauses.

## Where Altyst is the weaker choice

A comparison page where one column wins every row is an advertisement. So:

- An ARGUS file as a deliverable. Covered above, and for some readers it is the whole answer.
- The deepest lease-by-lease work. Altyst models rollover with TI, LC, downtime, renewal probability and expense stops, which covers the large majority of office, retail and industrial deals. For a lease with a genuinely unusual recovery or option structure, bring that lease and test it before you decide anything.
- A structure the engine has not anticipated. A participating mortgage, an odd ground lease reset, a promote that pays differently before and after a refinance. Export the workbook and build that leg in Excel, which is exactly what the export exists for.
- Certifications. We hold none.
- Certified interchangeability. If your value comes from every analyst in a 40-person team opening any model with identical training, that is a real thing you would be walking away from.

## Test it in one afternoon

Do not take the table above on faith, ours or anyone else's. Run a real deal you have already underwritten.

1. Upload the actual package, including the ugly scanned page. Extraction quality on a clean broker Excel file tells you nothing about extraction quality.
2. Check the extracted rent roll against your own tie-out. Look at unit count, down and model units, and concessions specifically.
3. Rebuild your debt: size on loan-to-value, on a minimum debt service coverage ratio, and on a debt yield, and check whether the constraint that binds is the one you expected.
4. Push one assumption you know is load-bearing and confirm the downside case moved with it.
5. Reprice to a target return and see whether the number it gives you is a bid you would actually make.
6. Export the workbook, open it in Excel, and change a cell on a live tab.
7. Read the Provenance tab as if you were the person on the investment committee who did not build the model.

If step 7 does not answer more questions than your current process would, none of the rest matters.

## What it costs

Plans are Individual $12, Professional $24, Team $99 a month, each with a published price per additional deal and a published monthly ceiling on new deals. The unit is one deal creation. Uploading documents, running the underwriting, editing assumptions, recomputing and exporting are never metered, so repricing a deal eleven times costs the same as repricing it once. Unused deal credits roll over. AI document reading draws on an allowance carried by each deal, included or additional, and when a deal has used its allowance the model, the edits and every export keep working, and your other deals are unaffected. There is no implementation fee, no annual commitment, and no quote to wait for. Full detail on pricing (/pricing).

Questions

Q: Is Altyst an ARGUS alternative?
A: For some of the work, yes. For one part of it, no. Altyst reads an offering memorandum, rent roll and T-12 and returns a full editable underwriting model in which every figure read out of a document is traced back to that document and the excerpt it came from, which is the part of the job that happens before a modeling environment can be used at all. What it does not do is produce an ARGUS file. If a lender, appraiser, JV partner or buyer has asked for one by name, that requirement does not go away, and the honest answer is to keep the license for the deals that carry it.

Q: Can Altyst import my ARGUS files?
A: No. It does not read ARGUS files and it does not import a firm's Excel template. Altyst builds the model from the underlying deal documents instead: the offering memorandum, the rent roll, the trailing twelve, the lease schedule and the debt quote. The practical pattern is to screen and underwrite in Altyst and carry the deals that survive into whatever your firm uses to close.

Q: How much does ARGUS cost, and how much does Altyst cost?
A: Altus Group does not publish a price for ARGUS Enterprise, so you request a quote, and the figures that circulate on third-party sites differ from each other by roughly an order of magnitude. Ask Altus directly for the all-in annual number including seats, implementation and training. Altyst publishes every price on its pricing page, along with each plan's monthly ceiling on new deals and the price of an additional deal. There is no implementation fee and no annual commitment.

Q: Can I trust the numbers, or is an AI doing the math?
A: AI is used only to read documents and propose values, each shown with its source and a confidence for a person to accept or overwrite. It never performs the calculation. Every financial result is computed by a deterministic engine in exact decimal arithmetic, so the same inputs always produce the same output. An evaluator in the test suite re-reads the exported workbook's own formulas and fails the build if any of them disagrees with the engine. Altyst holds no SOC 2 or ISO 27001 certification, and that is stated on the security page as well as here.

Q: Can I export the model to Excel?
A: Yes. The exported workbook carries live formulas on its core tabs: EGI from gross potential rent, other income, vacancy and credit loss, NOI from EGI and operating expenses, levered cash flow from unlevered cash flow and debt service, and an IRR over the cash-flow range. Some tabs are engine-computed values rather than formulas, and the workbook's Cover tab names which tabs recalculate and says the rest do not. There are six export surfaces in total: the workbook, a memo PDF, an editable PowerPoint deck, a deck PDF, a lender package and an offering summary.

Q: Does Altyst handle lease-by-lease commercial modeling?
A: Yes, for office, medical office, retail, industrial, flex and mixed-use, with a tenant schedule and rollover that carries TI, LC, downtime, renewal probability and expense stops. It also covers multifamily, single-family and short-term rental, self-storage, hotel, warehouse, and land or development. Those run on four adapters built for four different income models, per-unit residential, lease-by-lease commercial, development carry, and a departmental hotel model, with class-specific defaults and lease economics inside the commercial one rather than one model with different labels. For a lease with an unusual recovery or option structure, test that lease before you decide anything.

Q: What happens to my models if I cancel?
A: The workspace, the deals, the models and the exports stay accessible. Only new paid processing pauses. The exports are files that open without Altyst. Deleting a deal removes its documents and results and produces a receipt, and deleting a workspace does the same across every deal in it, with the receipt naming anything still pending.

ARGUS, ARGUS Enterprise and the ARGUS Intelligence Platform are trademarks of Altus Group Limited. Altyst is not affiliated with, endorsed by, or sponsored by Altus Group. ARGUS is named here only to describe how commercial real estate models are commonly built, and to answer a question buyers ask by name. Statements about ARGUS on this page reflect Altus Group's own published materials as of 6 August 2026 and may have changed since. Confirm anything that matters with Altus directly.

---

## Altyst vs building the underwriting model in Excel
Source: https://altyst.ai/vs/excel
Updated: 2026-08-06

This is not a choice between software and a spreadsheet. Altyst ships an Excel workbook with live formulas as one of its outputs. What you are actually choosing is who does the retyping, and whether six weeks from now anyone can say where a number came from.

### Keep building it in Excel when

- Your firm's template encodes years of house convention and your investment committee reads it on sight. Nothing imports institutional memory.
- The structure is unusual enough that no schema fits it. A spreadsheet has no schema, which is exactly why it can model a participating mortgage, an odd ground lease, or a one-off JV nobody has seen before.
- You underwrite a handful of deals a year and the retyping is not the bottleneck.
- The model has to be handed to someone who will rebuild half of it anyway.

### Bring in Altyst when

- The documents arrive as a PDF offering memorandum, an Excel rent roll, and a scanned T-12, and somebody is retyping all three.
- You want the rent roll and the trailing statement checked against each other before anything enters the model.
- You are repricing against a bid deadline and need the price that hits your return, not another pass through the model by hand.
- Six weeks later, at the investment committee, you need to answer where the insurance line came from.
- You still want the workbook at the end. You get one, and its core tabs recalculate.

## Start with what the spreadsheet is good at

Anyone selling you software will skip this section. It is the most important one, because if you do not know what you are giving up you cannot tell whether the trade is worth it.

A spreadsheet has no schema. That single property is why real-estate finance still runs on it. Any structure a deal can have, a spreadsheet can hold: a participating mortgage, an earnout on a partial condemnation, a ground lease with a reset tied to an index nobody else uses, a promote that pays differently before and after a refinance. Software that owns your model has to have anticipated the structure. A blank grid never has to anticipate anything.

Your firm's template is also more than arithmetic. It is where the house convention lives: which line items get their own row, how you treat a tax reassessment on transfer, what your investment committee expects on page one. That is institutional memory, and no import wizard has ever moved it.

And the file outlives everything. An .xlsx opens in other spreadsheet programs, sits in a data room, and goes to a lender who expects a workbook and not a login.

Hold on to all of that. The case for anything else has to be made against it, not around it.

## Where the spreadsheet actually costs you

Four places, and none of them are about the math being wrong.

The retyping. A deal package arrives as a PDF offering memorandum, an Excel rent roll on a broker's own layout, a scanned trailing twelve, and a debt quote in an email. Every figure that reaches your model gets there because a person read it off one document and typed it into another. That is where the transposition happens, and it happens on the numbers that matter, because the numbers that matter are the ones there are the most of.

The tie-out that never gets done. The rent roll says one thing about in-place income and the T-12 says something slightly different, and reconciling them by hand takes an afternoon nobody has on a deal that is probably a pass. So it does not get done, and the discrepancy that would have killed the deal in week one survives into the bid.

Where a number came from. Open a model six weeks after you built it and look at the insurance line. The cell holds a value. It does not hold the page of the loss run it came from, whether it was the broker's number or the quote you got, or how confident you were. You reconstruct it from memory, or you email the analyst, or you accept it.

The fifth reprice. The first sensitivity run is fine. It is the fifth one, at nine at night, after the seller's counter, that finds the hardcoded number in the middle of a formula chain and the link to the workbook version that got renamed.

## What happens to the same package in Altyst

Documents in, an editable model out, and the arithmetic done by an engine rather than predicted by a language model.

- It reads the offering memorandum, rent roll, T-12, debt quote and lease schedule in PDF, Excel, CSV, Word, plain text, or a photograph of a page. Scanned pages go through OCR.
- Extracted values arrive as proposals, each shown with its source and a confidence, for you to accept or overwrite before it enters the model. The documents are also tied out against each other, and conflicts come to you to settle rather than being silently resolved.
- Every financial result is computed by a deterministic engine in exact decimal arithmetic. Not by a model that predicts a plausible number. The same inputs always produce the same result, which is the only property that makes a downstream audit possible at all.
- Every extracted figure traces back to the document it came from and the excerpt it was read from, with a page number when that excerpt can be placed on exactly one page, and you can click a number to see the formula underneath it.
- Change any assumption and returns, cash flow, debt, and the downside case recompute together.

## The part people do not expect: you still get the workbook

Altyst exports an Excel workbook, and its core tabs carry live formulas rather than pasted values. EGI is computed as gross potential rent plus other income less vacancy and credit loss. NOI is EGI less operating expenses. Levered cash flow is unlevered cash flow less debt service. The IRR row is an actual IRR over the cash-flow range. Open it in Excel, change a cell, and the sheet recalculates the way a model you built would.

Every workbook carries Cover, Model, Cash Flow, Debt, Scenarios, Sensitivities, Provenance and Model Checks. A deal picks up more depending on what it needs, among them Sources and Uses, Reprice, Partnership, After-Tax, Rent Roll, Tenant Schedule, Risks and Diligence. The Provenance tab works at the level of the assumption rather than the cell: every driving assumption, its value, whether it was your figure or one the system proposed, what kind of source stands behind it, and how firm that source is, plus any notes on how the documents were read. Document-and-excerpt attribution for a specific extracted figure lives on screen, not on that sheet.

Two details about that export, because they are the kind of thing you should be asking any vendor:

1. Not every tab recalculates. Some are computed by the engine and written in as values. The Cover tab names which tabs carry live formulas and says plainly that the rest do not recalculate, so nobody discovers it by editing a cell and watching nothing move. That line is generated by scanning the finished file for formula cells rather than typed by hand, so it cannot say a tab is live when its numbers are pasted.
2. The formulas are checked, not assumed. A separate evaluator in our test suite re-reads the workbook's own formulas, computes them, and compares the result against the engine. A workbook that wrote a formula disagreeing with the model fails the test run.

Ask every other vendor on your list for a sample export before you buy, and open it. Whether the cells contain formulas or numbers is the fastest single test of how seriously a product takes the handoff back to you.

## Side by side, on the things that decide it

| The job | Building it yourself in Excel | Altyst |
| --- | --- | --- |
| Getting figures out of a PDF | A person reads and retypes | Extracted, with source and confidence, for your review |
| Rent roll against the T-12 | Manual, so often skipped | Tied out, conflicts raised to you |
| Where a figure came from | The cell holds a value | Document, excerpt and confidence on screen; every driving assumption and its source on the Provenance tab |
| Changing an assumption | Recalculates if the links survived | Whole model recomputes, including the downside case |
| Two-way sensitivity grid | Build the table, wire the driver | Sweep the drivers, grid comes back |
| Repricing to a target return | Solve it by hand | Reprice to the bid that hits your return |
| The tenth deal this month | The tenth copy of the template | The same model shape every time, comparable across deals |
| A structure nobody has modeled | The blank grid wins | Export the workbook and build that leg in Excel |
| The file you hand a lender | The workbook | The workbook, plus a memo, a lender package, and a deck |
| Who owns it afterwards | You | You. The exports are yours and they open without us |

## What to test before you switch anything

Do not take the table above on faith, ours or anyone's. Run one real deal you have already underwritten by hand, and compare.

1. Upload the actual package, including the ugly scanned page. Extraction quality on a clean broker Excel file tells you nothing.
2. Check the extracted rent roll against your own tie-out. Look specifically at unit count, down and model units, and concessions.
3. Rebuild your debt: size on LTV, on minimum DSCR, and on debt yield, and see whether the constraint that binds is the one you expected.
4. Push one assumption you know is load-bearing, and confirm the downside case moved with it.
5. Export the workbook and open it in Excel. Change a cell on a live tab. Watch what recalculates.
6. Read the Provenance tab as if you were the person on the investment committee who did not build the model.

If step 6 does not answer more questions than your own spreadsheet would, none of the rest matters.

## What it costs, and what is metered

Plans are Individual $12, Professional $24, Team $99 a month, with a per additional deal price on each. The unit is one deal creation. Uploading documents, running the underwriting, editing assumptions, recomputing, and exporting are never metered, so repricing a deal eleven times costs the same as repricing it once. Unused deal credits roll over. AI document reading and property research draw on an allowance carried by each deal, included or additional, so one document-heavy deal never leaves a later one short, and when a deal has used its allowance the model, the edits, and every export keep working. The full detail, including each plan's monthly ceiling on new deals, is on Pricing (/pricing).

Questions

Q: Does Altyst replace Excel?
A: No, and it is not trying to. Altyst exports an Excel workbook with live formulas across its Model, Cash Flow, and Debt tabs, so the workbook is one of the deliverables rather than something you give up. What it replaces is the retyping of figures out of a PDF into a spreadsheet, and the loss of any record of where those figures came from.

Q: Can I keep using my firm's own Excel template?
A: Yes. Nothing stops you, and for a deal with an unusual structure it may still be the right tool. The common pattern is to screen and underwrite in Altyst, export the workbook, and carry it into your template for the deals that survive the screen. Altyst does not import a firm template or reproduce its layout.

Q: Do the exported Excel formulas actually work, or are they pasted numbers?
A: They are real formulas on the tabs that carry them: EGI from gross potential rent, other income, vacancy and credit loss, NOI from EGI and operating expenses, levered cash flow from unlevered cash flow and debt service, and an IRR over the cash-flow range. Some tabs are engine-computed values rather than formulas, and the workbook's Cover tab names which tabs recalculate and says the rest do not. An evaluator in the test suite re-reads the workbook's formulas and checks that they reconcile to the engine.

Q: Is the AI doing the math?
A: No. AI is used only to read documents and propose values, each one shown with its source and confidence for a person to review. Every financial result is computed by a deterministic engine in exact decimal arithmetic, so the same inputs always produce the same output. That separation is what makes the model auditable.

Q: What if the engine does not support the structure my deal needs?
A: Then you export the workbook and build that leg in Excel, which is exactly what the export exists for. Before you commit, run the structure you care about on the sample deal or on one of your own, and check it against how you model it today. A vendor that cannot show you the structure working on your deal has not answered the question.

Q: Where does my model live, and can I leave with it?
A: Deals and documents live in your workspace. The exports are files: an Excel workbook, a memo PDF, an editable PowerPoint deck, a deck PDF, a lender package, and an offering summary. They open without Altyst and they are yours. Deleting a deal removes its documents and results and produces a receipt. Deleting a workspace does the same across every deal in it, and its receipt names anything still pending. Both are described on the Security page.

Microsoft and Excel are trademarks of Microsoft Corporation. Altyst is not affiliated with, endorsed by, or sponsored by Microsoft. Excel is named here only to describe how underwriting models are commonly built, and how the workbook Altyst exports is meant to be used.

---

## How to evaluate real estate underwriting software
Source: https://altyst.ai/vs/underwriting-software
Updated: 2026-08-06

Most underwriting tools demo well, because a demo is run on a clean document by someone who knows where the product is strong. These are the questions that separate them, the answers worth accepting, and where we land on each, including the ones we fail.

### Decide which shape of product you are buying

- Some tools sit on top of the model you already have. They read documents into your workbook, validate it, and route it for approval. Your template stays authoritative, and so does its maintenance.
- Others own the model. They compute it themselves and hand you an export. You give up template control and get consistency across deals and a real audit trail.
- These are different purchases with different failure modes. Comparing one of each on a single feature list produces a meaningless score.
- Altyst is the second kind. If your firm's template is genuinely non-negotiable, say so on the first call and save everyone the cycle.

### Then take the same twelve questions to everyone

- Ask each vendor to run YOUR deal package, including the scanned page, not their demo file.
- Ask what computes the arithmetic, and ask them to run the same deal twice.
- Ask for a sample Excel export before you sign, and open it.
- Ask what happens to your models if you cancel.
- Ask what they do not do. A vendor with no answer has not thought about it, or is not telling you.

## The demo is not the test

Every underwriting product looks capable in a demo, because the demo runs on a document the vendor chose. The offering memorandum is text-native, the rent roll has one row per unit and no merged cells, and the presenter knows which button to avoid. Nothing you learn there survives contact with a scanned T-12 that came out of a copier at an angle.

So the useful evaluation is not a feature checklist. It is twelve questions, asked of everyone, with your own deal package on the table.

## The twelve questions

### 1. What actually computes the numbers?

Ask directly whether the financial results come from a deterministic calculation engine or from a language model, and then ask them to run the same deal twice and show you both results. If a returned IRR moves between two runs on identical inputs, nothing downstream of it can be audited: not the memo, not the committee deck, not the answer you give a lender.

Where Altyst lands: a deterministic engine in exact decimal arithmetic. AI reads documents and proposes values, and never performs the calculation. Same inputs, same result, every time.

### 2. Can you click a figure and see where it came from?

Not "we cite sources" as a category. Click a specific number, and ask what appears. A defensible answer names the document and shows you the text the figure was read from, tells you how sure it is, and reaches the exports rather than living only on screen.

Where Altyst lands: on screen, every extracted figure carries the document it came from, the excerpt it was read from, a confidence, and a page number when that excerpt can be placed on exactly one page. An excerpt that appears on several pages is left unpinned rather than guessed at, and a figure you typed yourself has nothing to trace. In the exported workbook there is a Provenance tab, and it works one level up: every driving assumption, whether it was your figure or a proposed one, what kind of source stands behind it, and how firm that source is.

### 3. What does it read, and what does it do when the page is a photograph?

Ask for the format list, then ask about the scanned page specifically. Ask what happens when the rent roll and the trailing statement disagree: whether the conflict is raised for a human to settle, or resolved quietly by whichever document was parsed last.

Where Altyst lands: PDF, Excel, CSV, Word, plain text and photos of a page, with OCR for scanned documents. Extracted values are proposed with a confidence for review, and documents are tied out against each other with conflicts raised to you.

### 4. Is the asset class model real, or a relabeled multifamily model?

The tell is the vocabulary. An office model that talks about units instead of leases, or a hotel model with no ADR and no departmental expenses, is a multifamily model wearing a hat. Ask to see lease-by-lease rollover on an office deal and departmental expenses on a hotel.

Where Altyst lands: multifamily and single-family, office and medical office, retail, industrial, flex and warehouse, self-storage, mixed-use, hotel and hospitality, land and development, redevelopment and adaptive reuse, and ground-up hotel development. Those run on four adapters built for four different income models: per-unit residential, lease-by-lease commercial, land and development carry, and a departmental hotel model on occupied room nights and ADR. Inside the commercial one, each class carries its own defaults and lease economics, so self-storage recovers no operating expenses and books no TI or leasing commissions while office recovers most of them and carries months of downtime. An office deal is not a multifamily model with the labels changed, but eight of those classes do share one lease-by-lease engine, and you should ask us to prove the difference on your own rent roll.

### 5. How does it size debt?

Three sizing tests decide most deals: loan to value, minimum debt service coverage, and debt yield. Ask which are supported, and whether the product tells you which constraint binds. Then ask about the structures you actually see: a mezzanine piece, an interest-only period, a bridge loan to a refinance.

Where Altyst lands: sizing on LTV, minimum DSCR, or debt yield, with mezzanine, interest-only, additional tranches, and bridge-to-refinance structures.

### 6. What does the rollover model carry?

For anything with commercial leases this is the whole model. Tenant improvements, leasing commissions, downtime between leases, renewal probability, and expense stops. A product that models expiries but not downtime overstates income on any deal with rollover inside the hold, because the months a suite sits empty never reduce anything.

Where Altyst lands: lease-by-lease rollover with TI, LC, downtime, renewal probability, and expense stops. Ask us the same follow-up you ask everyone: bring a rent roll with a messy expiry schedule and look at what the model does with it.

### 7. Is the waterfall real?

Ask for a preferred return that compounds, return of capital, a general partner catch-up, a residual promote, and multiple IRR hurdle tiers. Then ask what happens to a structure that does not fit, because something in your book will not fit.

Where Altyst lands: all of it, off by default. Structures outside that shape are a job for the exported workbook, which is the honest answer rather than a roadmap date.

### 8. Is after-tax an overlay or an afterthought?

If your investors care about after-tax returns, ask for depreciation on the correct schedule, recapture at sale, passive-loss carryforward, and a 1031 deferral, and ask whether the after-tax figures flow into the exports or only appear on one screen.

Where Altyst lands: an overlay covering straight-line depreciation, passive-loss carryforward, unrecaptured section 1250 gain plus capital gains at sale, and an optional 1031 deferral, on rates you supply. Bonus depreciation is in there too, and the cost-segregated slice it creates is treated as section 1245 property, so it recaptures at the ordinary rate rather than the section 1250 rate. The whole overlay is off until you turn it on, because a default tax rate is a wrong tax rate.

### 9. What comes out, and does the Excel contain formulas?

Ask for a sample export before you buy, and open it. A workbook of pasted values is a report. A workbook of live formulas is a model you can keep working in. Ask which tabs recalculate, and be suspicious of a vendor who does not know off the top of their head.

Where Altyst lands: an Excel workbook on live formulas, a one-page investment memo PDF, an editable PowerPoint deck, a deck PDF, a lender package, and an offering summary. Some workbook tabs are engine-computed values; the Cover tab names which tabs recalculate and says so plainly.

### 10. What happens if you leave?

Ask what you can take with you, in what format, and what happens to your deals after a cancellation. The good answer is that the exports are ordinary files and the work stays reachable.

Where Altyst lands: exports are files that open without us. After a cancellation the workspace, deals, models and exports stay accessible; only new paid processing pauses. Deleting a deal removes its documents and results and produces a receipt; deleting a workspace does the same across every deal in it, and its receipt names anything still pending.

### 11. What is the security posture, stated in specifics?

Ask for the subprocessor list by name, where documents are stored, whether uploads are used to train anyone's models, how deletion works, and which certifications they hold. On that last one, listen for the word "compliant", which is not an audit. A vendor with no certification who says so is more trustworthy than one implying an attestation they do not have.

Where Altyst lands: workspace isolation enforced server side, role-based access, encryption in transit and at rest, published subprocessors, documents never sold and never used to train third-party models, and no SOC 2, ISO 27001, or comparable certification. We say that last part on the Security (/security) page rather than leaving you to discover it in procurement.

### 12. What is the pricing shape, and what is metered?

The headline number matters less than the meter behind it. Ask what counts as a unit: a seat, a deal, a document, a run. Ask what happens when you hit a limit, whether unused capacity carries forward, and whether re-running a model costs anything.

Where Altyst lands: plans are Individual $12, Professional $24, Team $99 a month, with a stated per additional deal price and a published monthly ceiling on new deals for each plan. The unit is one deal creation. Recompute, edits and exports are never metered, and unused deal credits roll over. Details on Pricing (/pricing).

## A scorecard you can take to every vendor

Score each vendor on your own deal, not the demo file. The last three rows are the ones that get skipped in a demo and then decide the renewal.

| Criterion | What a good answer looks like |
| --- | --- |
| Computation | Deterministic engine, identical output on a repeat run |
| Provenance | The document and the text a figure was read from, on screen, and a provenance sheet in the export |
| Extraction | Your scanned page, not their sample, with confidence shown |
| Conflicts | Raised to a human, not silently resolved |
| Asset class | Vocabulary and mechanics native to the class |
| Debt | LTV, DSCR and debt yield, and it names the binding constraint |
| Rollover | TI, LC, downtime, renewal probability, expense stops |
| Waterfall | Compounding pref, return of capital, catch-up, promote, hurdles |
| Tax | Depreciation, recapture, carryforward, and it reaches the exports |
| Exports | Excel with live formulas, and they can say which tabs |
| Exit | Files you keep, work reachable after cancellation |
| Security | Named subprocessors, honest certification status |
| Pricing | The meter is legible, and re-running is free |

## Two things worth ruling a vendor out over

A number that changes between runs. If the same inputs produce a different IRR twice, the product cannot support an audit trail no matter what its marketing says about traceability. It costs you one repeat run to check, and it is the only item on this page you cannot be talked out of.

A refusal to show a sample export. The export is the handoff back to you. A vendor unwilling to send one before a contract is telling you something about what is in it.

## Where this leaves us

Altyst is the kind of product that owns the model. That is a real cost to you: you do not get to keep your firm's template as the authoritative artifact, and if a structure falls outside the engine you finish it in the exported workbook. What you get for it is that the arithmetic is deterministic and testable, every extracted figure carries the document and the excerpt it was read from, and the same model shape comes out of every deal so the tenth one is comparable to the first.

If that trade is wrong for your firm, one of the tools that sits on top of your own spreadsheet is the better purchase, and you should go make that call quickly rather than sitting through six demos to arrive at it.

Questions

Q: What is real estate underwriting software?
A: It is software that turns a deal's documents and assumptions into a financial model: income and expenses, debt, cash flow, returns, and the exports an investment committee or a lender expects. Products in the category differ mainly in whether they read the documents for you, whether they compute the model themselves or sit on top of a spreadsheet you maintain, and whether any figure can be traced back to the document it came from.

Q: How do I test extraction quality fairly?
A: Use your own package rather than the vendor's sample, and include the worst document in it: the scanned trailing statement, the rent roll with merged cells, the page photographed at an angle. Then check the extracted values against your own tie-out on the things that decide a deal, which are unit count, down and model units, concessions, and the operating expense lines that were restated.

Q: Does it matter whether the calculations come from an AI model?
A: It matters more than any other single answer. If a language model produces the financial results, the same inputs can produce different numbers on different runs, which means no audit trail is possible and no export can be reconciled. AI is well suited to reading a document and proposing a value for review. The arithmetic should come from a deterministic engine.

Q: What should I ask about security in a procurement review?
A: Ask for the subprocessor list by name, where documents are stored and for how long, whether uploads are used to train third-party models, how workspace isolation is enforced, what deletion actually removes, and which certifications the vendor holds. Treat the word compliant as a warning rather than an answer, and prefer a vendor who names what they do not have.

Q: Does Altyst have SOC 2 or ISO 27001?
A: No. Altyst holds no SOC 2, ISO 27001, or comparable third-party certification, and the Security page says so directly. What is in place is workspace isolation enforced server side, role-based access, encryption in transit and at rest, published subprocessors, and deletion that removes the documents and results and produces a receipt naming anything still pending.

Q: How long does an evaluation take?
A: One real deal is usually enough to separate a shortlist. Run the same package through each product, compare the extracted values against your own tie-out, push one load-bearing assumption and watch what recomputes, then open every sample export. Vendors who cannot handle your ugliest document, or who will not send an export, remove themselves from the list at that point.

---

## Altyst vs an analyst or an outsourced underwriting desk
Source: https://altyst.ai/vs/outsourced-underwriting
Updated: 2026-08-06

A person brings judgment, which no engine has. Software brings iteration and consistency, which no person can sustain across a full pipeline. Most firms need both, and the answer usually falls out of how many packages arrive in a month.

### Send it to a person when

- The deal has already survived the screen and the money is real.
- The question is judgment, not arithmetic: whether the submarket supports the rent, whether the sponsor is credible, what the seller will actually take.
- Something in the structure has no precedent, and someone has to decide how it should be modeled at all.
- You need a human to defend the assumptions in a room, out loud, under questioning.

### Do it in software when

- Twenty packages arrived this month and eighteen of them are a pass. Paying for judgment on the eighteen is the expensive mistake.
- You are iterating. The fifth reprice after the seller's counter costs nothing in software and costs a favor from a person.
- You want the same model shape on every deal, so this week's is comparable to last quarter's.
- The seller's package is confidential enough that fewer parties is better.
- You want the model in your hands, not in someone else's file.

## The two things being compared are not the same thing

An analyst, in house or outsourced, does two separable jobs. One is mechanical: read the package, tie the rent roll to the trailing statement, build the pro forma, size the debt, produce the memo. The other is judgment: decide whether the rent growth in the sponsor's model is defensible, whether the expense ratio is credible for that vintage in that submarket, whether the story holds.

The first job is where the hours go. The second is what you are actually paying for.

Software does the first job well and does not do the second at all. Nothing in Altyst has an opinion on whether a submarket supports the rent. It computes exactly what you tell it, the same way every time, and shows you where each input came from. That is the whole product. A comparison that blurs that boundary is not worth reading.

## Turnaround, and what it does to a pipeline

The constraint on a person is a queue. Deals arrive in bursts, and the queue lengthens exactly when the market is busiest, which is exactly when a bid deadline is shortest. The constraint on software is your own attention.

That difference matters most at the top of the funnel. A firm looking at twenty packages a month passes on most of them. Every hour of skilled human time spent on a deal that was always going to be a pass is a real cost, and it is invisible because it never shows up next to a deal that closed.

## Iteration, which is where the cost hides

Underwriting is not one calculation. It is a first pass, then the exit cap the investment committee actually believes, then the debt quote that came back different, then the seller's counter, then the version where you assume the tenant does not renew.

Each of those is a round trip with a person. With software, changing an assumption recomputes returns, cash flow, debt and the downside case together, and repricing to the bid that hits your return is a single operation rather than a request.

This is also where the meter matters. In Altyst the unit is one deal creation. Recompute, edits and exports are never metered, so the eleventh reprice costs exactly what the first one did. Plans are Individual $12, Professional $24, Team $99 a month; the full detail is on Pricing (/pricing).

## Consistency across a book

Two analysts build two different models. So does one analyst in March and the same analyst in November. The line items drift, the vacancy convention drifts, the exit assumption gets a new treatment on a deal where it mattered, and by the time you want to compare this quarter's pipeline against last year's you are comparing artifacts, not deals.

An engine does the same thing every time. That is a limitation when the deal is unusual and an advantage across a pipeline, and it is the reason the comparison view across deals is meaningful at all: the numbers were produced the same way.

## Confidentiality, plainly

A seller's package with a rent roll in it is confidential information about people's tenancies. Every additional party who handles it is an additional place it exists. That is not an accusation about any service provider. It is arithmetic.

Where documents go in Altyst is published rather than described: they live in your workspace, the subprocessors are named on the Legal (/legal) pages, uploads are not sold and are not used to train third-party models, and deleting a deal removes its documents and results and produces a receipt, with a workspace deletion doing the same across every deal in it and naming anything still pending on the receipt. Whatever you use, ask for the same specifics in writing.

## Who ends up holding the model

At the end of an outsourced engagement you have a file. At the end of a software workflow you have a model you can open, change, and re-run, plus the exports: an Excel workbook on live formulas, a one-page memo, an editable deck, a lender package, and an offering summary.

The difference shows up the week after, when the investment committee asks for the version where the renovation is phased over two years instead of one.

## Where we would draw the line

Use software for the funnel and a person for the decisions. Screen everything in Altyst, so that the deals that reach a human have already been tied out, modeled, stressed, and priced, and the human hours go to the handful that deserve them.

If you have no in-house analyst at all, software gets you a defensible model and the documents an investment committee expects. It does not get you someone to argue with about the rent comps, and you should not buy it expecting that.

Questions

Q: Can underwriting software replace an acquisitions analyst?
A: It replaces the mechanical half of the work: reading the package, tying the rent roll to the trailing statement, building the pro forma, sizing debt, and producing the memo. It does not replace judgment about a submarket, a sponsor, or a story, and it has no opinion about whether an assumption is realistic. Most firms use software to screen the pipeline and human time on the deals that survive.

Q: What is the real cost difference?
A: The visible cost is a subscription against hours. The hidden cost is iteration: with a person, every reprice is another round trip, and with software it is free. In Altyst the metered unit is deal creation, so recomputing, editing and exporting a deal cost nothing beyond it. Current plan prices are published on the Pricing page.

Q: How does this compare on turnaround?
A: A person is a queue, and the queue is longest exactly when the market is busy and bid deadlines are short. Software has no queue, so the throughput limit is your own attention. That matters most at the top of the funnel, where most packages are a pass and human hours spent on them never show up as a cost.

Q: Is it safe to put a seller's confidential package into software?
A: Ask for specifics from any party you send it to, human or software. For Altyst: documents live in your own workspace with isolation enforced server side, subprocessors are published by name, uploads are not sold and are not used to train third-party models, and deleting a deal removes its documents and results and produces a receipt. A workspace deletion does the same across every deal in it, and its receipt names anything still pending. Altyst holds no SOC 2 or ISO 27001 certification, which the Security page states directly.

Q: What do I actually get at the end?
A: A live model you can reopen and change, plus exports: an Excel workbook on live formulas, a one-page investment memo PDF, an editable PowerPoint deck, a deck PDF, a lender package, and an offering summary. Every figure in them reconciles to the model, and every extracted figure carries the document and the excerpt it was read from.

# Boundaries

- Altyst produces model outputs for screening and analysis. It is not investment,
  legal, tax, appraisal, or brokerage advice, and it does not guarantee accuracy or returns.
- No language model computes a financial result.
- No SOC 2, ISO 27001, or comparable third-party certification is held or claimed.
- Uploaded documents are not sold and are not used to train third-party models.
- There is no free tier, no annual billing, and no per-deal purchasing without a plan.
- Only public marketing pages are indexable. Everything under /deals, /account, and /team is
  a signed-in application surface and is excluded in robots.txt.