How to underwrite a multifamily deal

Everything in an offering memorandum is true in the narrow sense and shaped in the broad sense. The rents are real rents, just not the ones being collected today. The expenses are real expenses, just not all of them. Your job is to take the same property and produce a different, defensible set of numbers, then decide what you would pay for the cash flows those numbers imply.

This walkthrough does that end to end on a 24-unit deal. Every figure ties. You can rebuild it in a blank spreadsheet in about ninety minutes, and if you are new to this, you should, once, by hand. After that, stop doing it by hand.

The deal

A 1984-vintage garden-style 24-unit in a secondary market. Two-story walk-ups, surface parking, individually metered electric, master-metered water and sewer. Asking price $3,000,000, which is $125,000 per unit and $145.63 per square foot. The broker's flyer says "8.3% cap."

Annualized, that is $358,800 of in-place scheduled rent against $372,960 of rent at market. Twenty-two of twenty-four units are occupied, so physical occupancy is 91.7%. Hold both of those numbers. Almost every mistake in multifamily underwriting is a confusion between them.

Before anything else, get four documents. If you cannot get them, you cannot underwrite the deal, and you should say so out loud rather than modeling around the hole:

1. The rent roll, dated, with lease start and end dates, deposits, and balances owed.
2. The trailing twelve (T-12) operating statement, month by month, not a single annual column.
3. The last two years of operating statements, so you can see which expenses are lumpy and which are structural.
4. The tax bill and the assessor's record, so you know the current assessed value and the effective rate.

Step 1: gross potential rent

Gross potential rent (GPR) is every unit at market rent, twelve months, as if the property were perfectly full and every tenant paid asking.

GPR = 24 units at market rent x 12 months = $372,960

GPR is a deliberately fictional number. It exists so that every leakage below it can be stated as a percentage of the same base, which is what makes deals comparable.

The mistake to avoid. Some models start with in-place rent and call it GPR. Others start with market and never subtract the gap. Either convention works if you are consistent. Mixing them double-counts or vanishes the difference, and on this deal the difference is $14,160 a year, which is roughly $227,000 of value at a 6.25% cap.

Step 2: the four leakages

Four separate things stand between GPR and money in the bank. Model them as four lines, not one blended "vacancy" plug, because they have different causes and different fixes.

Loss to lease is the gap between market rent and the contract rent on units that are already leased. Here that is $1,180 a month, or $14,160 a year, on the 22 occupied units. The two vacant units have no loss to lease because they will be leased at asking, and their empty months are picked up by the vacancy line instead.

Vacancy is rent lost to empty units. Physical vacancy today is 8.3% (two of twenty-four). That is a snapshot, not a forecast. Underwrite the number you believe you can sustain, informed by the submarket and by the T-12's own history. We use 6.0%.

Concessions are the free weeks and gift cards used to fill units. They are rent you never collect but that never shows as vacancy, which is exactly why they belong on their own line. We use 1.0%.

Credit loss, also called bad debt, is rent billed to a tenant who does not pay it. Pull it from the T-12's write-off line and from the rent roll's balances-owed column. We use 1.5%.

That 87.7% is economic occupancy: the share of theoretical rent that becomes revenue. Physical occupancy is 91.7%. The four-point spread is the deal's real operating story, and it is invisible if you only look at the occupancy percentage on the front page of the offering memorandum.

Step 3: other income

Everything the property collects that is not rent. Be specific, and be careful about what already exists versus what someone hopes will exist.

Two rules. First, only underwrite income the property is collecting now, unless you can point at a signed change (a RUBS program already noticed to tenants, a laundry contract already executed). "The buyer could implement RUBS" is a business plan, not revenue. Second, tie every line to the T-12. If the offering memorandum says $28,000 of other income and the T-12 says $18,970, the offering memorandum is showing you a plan.

Step 4: effective gross income

EGI = net rental income + other income = $327,098 + $18,970 = $346,068

EGI is the top line that matters. Every expense ratio, every management fee, and every sanity check runs off it.

Step 5: rebuild the expenses

Do not accept the seller's expense column. Rebuild it line by line, then add the lines the seller does not have because the seller does not pay them.

Four of those lines are where deals are won and lost.

Property taxes. In most jurisdictions a sale triggers reassessment, and the new assessed value is your purchase price. The seller bought in 2014 and is paying $31,500. At a 1.60% effective rate on a $3,000,000 sale, you will pay $48,000. That single line is a $16,500 swing in NOI, worth about $264,000 of value at a 6.25% cap. Check your state: a few cap annual increases or reassess on a cycle rather than on transfer, and some abate for a period. Get the answer from the assessor, not from the broker.

Insurance. The seller's premium is priced off the seller's loss history and a policy written years ago. Get a real quote from a broker who has seen the loss runs. In wind, hail, and wildfire exposed states this line has moved more than any other in recent years, and inheriting the old number is the fastest way to be wrong by six figures of value.

Payroll. The seller lives twenty minutes away and does the maintenance himself, so his statement shows no payroll. You will not do that. Price the labor at what it costs to buy: here, a part-time tech at roughly $18,000 fully burdened. If instead you plan to contract it all out, payroll goes to zero and repairs and maintenance goes up. What you cannot do is have neither.

Management. Third-party management on a 24-unit typically runs 4% to 8% of EGI, often with a per-unit monthly floor. We use 5.0%, which is $17,303, and we check it against a $50 per unit per month floor of $14,400. The percentage governs, so 5.0% stands. If you plan to self-manage, still charge the fee. Your time is not free, and a buyer at exit will underwrite the fee whether or not you did.

Step 6: NOI, and the reserves argument

NOI = EGI - operating expenses = $346,068 - $167,903 = $178,165

Now the question that trips up every junior analyst: do replacement reserves come out before or after NOI?

The convention that matters is this. Reserves sit below NOI for valuation and above the line for the lender. Cap rates in the market are quoted on NOI before reserves, so if you deduct reserves and then apply a market cap rate, you will systematically undervalue every deal. But agency and most bank lenders size debt on net cash flow, which is NOI less an underwritten replacement reserve, so your DSCR test has to use the lower number.

We carry $300 per unit per year, or $7,200. That is a routine-replacement figure. It does not cover the roofs and the parking lot, which are known, dated items and belong in the capital plan below.

Going-in cap rate on the asking price: $178,165 / $3,000,000 = 5.94%

The flyer said 8.3%. That gap is not a rounding error, and it is worth understanding precisely.

Step 7: where the broker's 8.3% came from

Six differences, each of them individually defensible as a marketing choice and collectively worth more than a third of the asking price.

Their NOI is $248,923 and an 8.30% cap on $3,000,000. Ours is $178,165 and a 5.94% cap. Neither statement is a lie. Only one of them is what the property will actually do under your ownership.

Most offering memoranda stack two or three of these. This one stacks all six, because every item is a real thing that appears in real packages and it is more useful to see the whole list at once. The companion piece, underwriting red flags in an offering memorandum, goes through each of them and how to catch it in the documents.

Step 8: your basis is not the price

You are not buying at $3,000,000. You are capitalizing everything it takes to own the asset on day one.

Cap rate on price = $178,165 / $3,000,000 = 5.94%

Cap rate on basis = $178,165 / $3,237,000 = 5.50%, also called yield on cost

The second number is the honest one. It is the return the asset produces on every dollar you have put into it, and it is the number to compare against your cost of capital. A deal that looks fine at a 5.94% going-in cap and requires $180,000 of immediate work is not a 5.94% deal.

Get the capital number from a property condition assessment, not from a walk-through. Roofs, parking lot, sewer laterals, panels, and windows are the five items that most often turn a good deal into a flat one.

Step 9: size the debt three ways

Lenders do not quote you a loan amount. They quote you constraints, and the smallest one governs. Assume a 6.25% fixed rate, 30-year amortization, 10-year term.

Test 1, loan to value. At 70% of the $3,000,000 price: 0.70 x $3,000,000 = $2,100,000

Test 2, debt service coverage ratio. The lender wants net cash flow to cover annual debt service by at least 1.25x. First compute the loan constant, which is annual debt service divided by loan amount:

Then work backwards:

Test 3, debt yield. Net cash flow divided by loan amount. It is the lender asking "if I foreclose tomorrow, what unlevered return do I own?" It has no rate and no amortization in it, which is why lenders trust it more than DSCR. At a 9.0% minimum: $170,965 / 0.09 = $1,899,611

Take $1,850,000. Annual debt service is $1,850,000 x 7.3886% = $136,689, or $11,391 a month. The resulting LTV is 61.7%, not 70%. This is the normal outcome in a higher-rate market: coverage binds before value does, and the equity check is larger than the LTV headline implies.

Equity required = $3,237,000 basis - $1,850,000 loan = $1,387,000

Step 10: negative leverage, and why year one looks bad

Compare two numbers: a going-in cap rate on price of 5.94% against a loan constant of 7.39%.

The debt costs more per dollar than the asset yields per dollar. That is negative leverage, and it means every dollar you borrow lowers your year-one cash return.

Against an unlevered current return on basis of 5.50%. The leverage did not help; it cost you three points of current yield in exchange for a claim on future growth.

Negative leverage is not automatically disqualifying. It is a statement that the entire return depends on rent growth, amortization, and the exit, none of which you control. It should raise your required margin of safety, not end the conversation.

Step 11: break-even occupancy

The single most useful stress number, and it takes one line:

You have underwritten 88.3%. The cushion is 8.7 points of occupancy before the property stops covering itself. Whether that is comfortable depends on the submarket's worst twelve months, which you can look up. State the convention you used, because some people compute break-even on rental income only and get a different number from the same deal.

Step 12: project the hold

Five-year hold. Three assumptions, stated plainly, because they are doing most of the work:

• Revenue grows 3.0% a year. All four leakage lines stay constant as a percentage of GPR.
• Operating expenses grow 3.5% a year, except management, which stays at 5.0% of EGI.
• Reserves grow 3.5% a year.

Expenses growing faster than revenue is deliberate. It is what has actually happened in most markets, and it is the difference between a model and a hope. Note the consequence: revenue grows 3.0% but NOI grows 2.6%, because a 48.5% expense ratio means a faster-growing denominator eats into the margin every year.

Year 6 is in the table for one reason: you need it to compute the exit.

Step 13: the exit

Three decisions, and they are worth more than everything above them combined.

The exit cap rate. You bought at a 5.94% going-in cap. In five years the building is five years older and you have no idea what the rate environment will be. Expanding the cap by 25 to 50 basis points over the hold is the standard conservative posture. We use 6.25%.

Trailing or forward NOI. Buyers capitalize the next twelve months of income, not the last twelve. So apply the exit cap to Year 6 NOI, not Year 5. On this deal that convention is worth $84,000 of gross value. Whichever you choose, write it down, and never compare an IRR built on forward NOI to one built on trailing.

Cost of sale. Broker commission, transfer taxes, legal, and closing prorations. 2% is a reasonable placeholder for a deal this size, and it is one of the few assumptions nobody will argue with.

The loan balance is not the loan amount. Five years of a 30-year amortization schedule retires $123,261 of principal, which is a real part of your return and one that people forget when they use interest-only shorthand.

Remaining balance = L x [(1+i)^360 - (1+i)^60] / [(1+i)^360 - 1] = $1,850,000 x 0.933373 = $1,726,739

Step 14: IRR and equity multiple

Line up the equity cash flows. Year 0 is what you put in; Year 5 includes both the year's operating cash flow and the net sale proceeds.

Equity multiple is the easy one. Total dollars out divided by total dollars in: $1,660,562 / $1,387,000 = 1.20x

IRR is the discount rate at which the present value of those inflows equals your Year 0 outflow. There is no closed-form solution, so it is found by iteration. In a spreadsheet it is =IRR() on annual flows or =XIRR() on dated ones. By hand, guess a rate, compute NPV, and interpolate:

The answer at the asking price: a 3.9% IRR and a 1.20x equity multiple over five years.

Two notes on IRR before you quote one. First, it is exquisitely sensitive to timing, so annual end-of-period flows and monthly flows on the same deal will not produce the same number. Say which you used. Second, IRR ignores size and the multiple ignores time, which is why you quote both. A 20% IRR on a 14-month flip and a 20% IRR on a seven-year hold are not the same business.

At 3.9% you are being paid less than a Treasury to take construction risk, tenant risk, rate risk, and illiquidity. That is not a negotiation. That is a no.

Step 15: what actually moves the answer

Before you walk away, find out which assumption you are really betting on. Change one input at a time and hold everything else. This table is run at the bid price established in the next section, so the base is 13.8%.

Read that table carefully, because it is the whole lesson of this article.

The two assumptions that move the answer most, the exit cap and rent growth, are the two you cannot verify. The one you can verify to the basis point, the interest rate, barely moves it. And the rate barely moves it here specifically because the loan was constrained by value rather than by coverage at the bid price, so a higher rate raised the payment without shrinking the proceeds. Change the price and that stops being true.

Anyone who tells you their model is precise is telling you they have not run this table.

Running these one at a time by hand means five saved copies of a spreadsheet and a real chance that one of them has a stale link. This is the part of the job worth automating: in Altyst you change the exit cap and every dependent figure recomputes at once, and each output opens into the formula and the inputs behind it, so you can see exactly which cell moved the IRR.

Step 16: solve backwards for the price

You do not walk away from a deal. You bid on it. The useful output of an underwriting is not "no," it is "yes at a number."

Set your target, here a 13% to 14% levered IRR over five years, and solve for the price that produces it. Remember that the price feeds back into three places: the loan amount, the acquisition costs, and the property tax line, since the assessment follows the sale.

At $2,600,000, which is $108,333 per unit and $126.21 per square foot:

Notice the constraint flipped. At $3,000,000 coverage bound the loan; at $2,600,000 value binds it. Always run all three tests. The binding constraint moves as the price moves, and assuming it stays put is how people over-promise proceeds to an equity partner.

That is your bid: $2,600,000, or 13.3% below ask. And now you can defend it line by line, which is a completely different conversation from "that feels high." Brokers do not respect low offers. They respect low offers with a rent roll analysis attached.

Note that even at the bid, year one is still slightly negatively levered: a 7.10% cap against a 7.39% constant. The return comes from growth, amortization, and the exit, not from day-one yield. Say that to your investors before they read it in a quarterly statement.

Two adjustments worth knowing

Interest-only. If the lender offers two years of interest-only, Year 1 debt service drops from $134,473 to $113,750 and cash-on-cash jumps from 4.2% to 6.3%. Nothing about the asset changed. Interest-only moves cash forward and leaves a larger balance at exit, which helps IRR through timing and hurts the equity multiple through the payoff. It is a financing choice, not performance, and presenting it as performance is one of the quieter dishonesties in this business.

Value-add. The base case above assumes you renovate nothing and earn no premium. If instead you spend $9,000 a unit on interiors across all 24 units and achieve a $125 per month premium:

A 14.5% return on cost is a good renovation program. But the premium is an assumption, not a fact, and it is the single most over-assumed number in multifamily. Prove it with leases signed at the renovated rent, in that submarket, in the last six months. Three signed comps beat any rent survey.

What this example leaves out

Say these out loud rather than letting them hide:

• Working capital. Most buyers fund a three-month operating reserve at close. Add it to equity and your IRR falls slightly.
• Sponsor fees. Acquisition, asset management, and disposition fees, if this is syndicated. They come out of the same cash flows.
• The partnership waterfall. Everything above is deal-level. Preferred returns and promotes split it, and the LP IRR is always below the deal IRR.
• Taxes. Depreciation, recapture, and the 1031 exchange change after-tax outcomes materially, and after-tax IRR on a levered deal typically lands well below pre-tax.
• Refinance. A five-year hold with a ten-year loan can be refinanced instead of sold.
• Monthly timing. Annual end-of-period flows are a simplification.

None of these are optional in a real underwriting. They are omitted here so the core arithmetic stays legible.

The order of operations, compressed

1. Rent roll first. Unit mix, in-place rent, market rent, lease expirations, balances owed.
2. GPR at market. Twelve months, every unit.
3. Subtract loss to lease, vacancy, concessions, credit loss. Four lines, not one.
4. Add other income that already exists.
5. EGI.
6. Rebuild every expense. Reassess taxes, re-quote insurance, add payroll and management.
7. NOI. Reserves below the line for value, above the line for the lender.
8. Basis, not price. Add closing, diligence, and immediate capital.
9. Size debt three ways. LTV, DSCR, debt yield. Take the lowest.
10. Compare the loan constant to the going-in cap. Know if you are negatively levered.
11. Break-even occupancy.
12. Five-year projection with expenses growing faster than revenue.
13. Exit on forward NOI, at a cap wider than your going-in, net of cost of sale.
14. IRR and equity multiple. Quote both.
15. Sensitivity table. Find out what you are actually betting on.
16. Solve backwards for the price that clears your hurdle. That is your bid.

Do this faster than by hand

Everything above is arithmetic. The reason it takes most analysts a day per deal is not the math, it is the transcription: pulling 24 rows off a PDF rent roll, keying twelve months of a scanned T-12, and then checking whether the two documents agree before any of it means anything.

That last step is the one people skip. If the rent roll annualizes to $358,800 and the T-12 shows $381,000 of rental revenue, something is wrong, and it is worth knowing that before you build a five-year model on top of it.

Altyst reads the rent roll and the operating statement together and ties them against each other, so figures that will not reconcile come back flagged with the document they came from rather than flowing quietly into your NOI. Then the model above runs on the extracted numbers, every output opens into the formula behind it, and the sensitivity table is a slider instead of five saved copies. Underwrite any property in seconds.

Try it on a deal you are already working on, then check its NOI against the one you built by hand. That is the right way to trust a tool.

Illustrative figures throughout. Expense levels, tax rates, insurance costs, and achievable rents vary widely by market and by property, and every one of them should be replaced with a quote or a comp before you sign anything. This is an educational walkthrough, not investment advice.