An acquisition model is a financial analysis framework that evaluates the purchase, financing, and projected returns of a real estate investment from entry through exit. It consolidates the purchase price, closing costs, debt structure, operating assumptions, and exit proceeds to calculate key return metrics like IRR and equity multiple for underwriting decisions.
Relevant Articles
- Need detail on the operating phase? See our guide to building a multifamily pro forma.
- Want to understand the mental model behind breaking down complex calculations? Review our Decomposition framework.
- Looking to model profit distributions at exit? See how to build a waterfall model.
Working Example: Project "Riverside"
To see how to build acquisition model ai in practice, let's model a specific value-add multifamily deal:
This example will anchor every calculation in this article. When AI generates formulas or logic, these specific numbers ensure consistency and prevent generic output.
Acquisition Model Components
An acquisition model contains three calculation layers: Sources & Uses (entry), Annual Operating Cash Flow (holding period), and Exit Proceeds (disposition). Most analysts mistakenly think of acquisition models as "pro formas," but the acquisition model is broader—it includes the capitalization structure and return calculations that sit outside the operating forecast.
The Sources & Uses block establishes total project cost and how it's financed. Uses include purchase price, closing costs, and any upfront capital expenditures deployed at acquisition. Sources list the debt proceeds and equity contributions. These must balance exactly. This is the first verification test: if Sources ≠ Uses, the model is broken before you even begin operations.
The Annual Operating Cash Flow section projects Net Operating Income (NOI) by year, subtracts debt service, and calculates cash available for distribution. This is where most of the detail lives, but it's not where the acquisition decision is made. Sponsors care about the IRR and equity multiple, which require both entry and exit values. For that reason, when learning to build acquisition model ai, you should start with the entry and exit blocks, not the operating detail.
The Exit Proceeds block calculates the gross sale price (stabilized NOI ÷ exit cap rate), subtracts selling costs and loan payoff, and delivers net proceeds to equity. This is the final cash distribution that feeds into the IRR calculation. In institutional models, this block also triggers the waterfall distribution if a promote structure exists, allocating proceeds between LP and GP based on IRR hurdles.
When prompting AI to generate an acquisition model, specify all three layers. If you ask for a "pro forma," you'll get operating projections but no debt structure or return metrics. If you ask for "returns analysis," you'll get IRR calculations with no backing detail. The acquisition model is the integration of all three. Structure your prompt to reflect this: "Build an acquisition model with Sources & Uses, 5-year operating forecast, and exit calculation assuming a 5.25% cap rate."
Decomposition (Purchase vs. Ops vs. Exit)
The biggest modeling error analysts make is mixing entry, operations, and exit logic into a single tangled block. This creates circular references, makes debugging impossible, and guarantees AI will hallucinate formulas. The solution is Decomposition—separating the model into three isolated calculation stages that flow in one direction only.
Start with the Purchase block. This section calculates total project cost (purchase price + closing costs + upfront CapEx) and determines equity required after debt proceeds. In Project Riverside, purchase price is $28,800,000, closing costs are 2.5% ($720,000), and upfront CapEx is $5,400,000. Total Uses equal $34,920,000. The loan is 65% LTV on total capitalization, yielding $22,698,000 in debt proceeds. Equity required is $13,050,000 (the difference). This block outputs a single number: equity invested. No other section should calculate this value.
Next, isolate the Operating block. This section takes stabilized NOI as an input (either modeled separately or hard-coded as an assumption) and calculates annual cash flow after debt service. In our model, Year 1 NOI is $1,620,000, debt service is $1,587,420 (assuming 5.5% interest, 30-year amortization), leaving $32,580 in distributable cash. By Year 5, after renovations stabilize, NOI reaches $2,340,000, and cash flow grows to $752,580. The Operating block does not reference the purchase price or exit value—it only converts NOI into cash available for distribution.
Finally, the Exit block calculates sale proceeds. Take Year 5 stabilized NOI ($2,340,000), divide by the exit cap rate (5.25%), and subtract selling costs (2% of gross sale price) and remaining loan balance. For Project Riverside, gross sale price is $44,571,429, selling costs are $891,429, and loan payoff is $21,240,000, leaving $22,440,000 in net proceeds to equity. This is the final cash distribution that, combined with annual cash flows, feeds into the IRR calculation.
When you decompose the model this way, each block can be verified independently. You can test whether Sources = Uses without worrying about exit logic. You can confirm debt service calculations without involving the purchase price. And when you prompt AI to "generate the Exit block," you provide it with a narrow, well-defined task that reduces hallucination risk.
This is the core principle of Decomposition: break the model into sequential, non-overlapping calculation zones. AI performs best when the instruction scope is constrained. If you ask ChatGPT to "build a full real estate model," it will generate a 200-row jumble. If you ask it to "calculate exit proceeds given $2,340,000 NOI, 5.25% cap rate, 2% selling costs, and $21,240,000 loan balance," it will return a clean, verifiable formula.
Purchase Price and Closing Costs
The Sources & Uses block is the entry point of every acquisition model. It answers a single question: How much equity must the sponsor raise to close this deal? Many analysts skip this section or treat it as an afterthought, but institutional investors review it first—they need to know the total check size and leverage ratio before evaluating returns.
Start with Uses. The minimum detail includes purchase price, closing costs, and upfront capital expenditures. For Project Riverside, the purchase price is $28,800,000 (the negotiated contract price). Closing costs are typically 2-3% of purchase price and include title insurance, legal fees, lender fees, and transfer taxes. We assume 2.5%, or $720,000. Upfront CapEx is the immediate capital required to execute the business plan—in this case, $5,400,000 to renovate 180 units at $30,000 per unit. Total Uses: $34,920,000.
Next, define Sources. The two sources are debt and equity. Debt is sized as a percentage of total project cost (Loan-to-Cost) or as a percentage of stabilized value (Loan-to-Value). In this example, the lender provides 65% LTV on total capitalization, resulting in $22,698,000 in debt proceeds. The remaining $13,050,000 must come from equity. This is the number that appears in the investor deck as "equity required."
A common mistake when learning to build acquisition model ai is failing to specify how debt is sized. If you prompt "include a loan" without providing a constraint, AI will hallucinate an arbitrary loan amount. The correct prompt is: "Size the loan at 65% LTV on total project cost, assuming a 5.5% interest rate and 30-year amortization. Calculate the loan amount, then derive required equity as the difference between total Uses and debt proceeds."
Another error is double-counting costs. Closing costs should only appear in the Uses section—do not subtract them from the purchase price elsewhere. And CapEx deployed at entry should not reappear in the operating forecast. When you isolate the Purchase block, these errors become obvious. If total Uses don't balance to total Sources, the model cannot proceed.
To verify this section, run the balance test: Sources − Uses = 0. If the result is non-zero, either the loan formula is incorrect or an equity contribution is missing. This is the first gate in the model—nothing downstream is valid until this test passes. In institutional models, the Sources & Uses block also includes a contingency reserve (2-5% of CapEx) and a debt service reserve (3-6 months of payments). These ensure the sponsor has liquidity to handle construction delays or lease-up shortfalls.
When prompting AI, provide all Uses components explicitly: "Calculate total Uses as the sum of $28,800,000 purchase price, 2.5% closing costs, and $5,400,000 upfront CapEx. Then calculate debt at 65% LTV and equity as the residual." This level of specification prevents the model from omitting closing costs or miscalculating equity.
Operating Assumptions
The Operating section of an acquisition model is where most analysts waste time over-engineering projections. For acquisition decisions, you do not need monthly rent rolls, 40-line expense budgets, or probabilistic vacancy assumptions. You need a defensible stabilized NOI and a rough path to get there. This is scope discipline—you're underwriting a purchase decision, not managing the asset.
For Project Riverside, the operating assumptions are intentionally sparse. Year 1 NOI is $1,620,000 (based on T-12 trailing performance). By Year 3, after renovations are complete, NOI stabilizes at $2,340,000—a 44% increase driven by rent growth and occupancy improvements. This is the only operating detail required to calculate returns. If you're tempted to model line-item expenses or lease-by-lease rent growth, you're building a pro forma, not an acquisition model.
The distinction matters when prompting AI. If you ask for "a detailed operating forecast," you'll get 30 rows of revenue and expense line items, most of which are irrelevant to the acquisition decision. The correct prompt is: "Generate a 5-year NOI projection starting at $1,620,000 in Year 1 and growing to $2,340,000 by Year 3, then holding flat. Do not model individual revenue or expense lines."
From NOI, subtract annual debt service to calculate Cash Flow Before Promote. Debt service is a constant payment calculated using the loan amount ($22,698,000), interest rate (5.5%), and amortization period (30 years). The annual payment is $1,587,420. In Year 1, this leaves $32,580 in cash flow. By Year 5, cash flow grows to $752,580 as NOI stabilizes and principal paydown reduces the interest portion of debt service.
This is the second verification test: does Cash Flow = NOI − Debt Service for every year? If the formula references anything other than these two inputs, the model is wrong. One error we see constantly is analysts adding back principal repayment to cash flow, double-counting it as both a return of equity and a distribution. Principal paydown is already reflected in the loan balance reduction, which increases net proceeds at exit. Do not add it to annual cash flow.
For acquisition models, the operating section should be a single table: Year, NOI, Debt Service, and Cash Flow. If your model exceeds this, you're over-specifying. The purpose of this block is not to predict operations with precision—it's to convert a stabilized NOI assumption into an equity IRR. The multifamily pro forma is where you build the operating detail. The acquisition model references the output of that pro forma (stabilized NOI) but does not recreate it.
When teaching AI to build this section, specify the flow explicitly: "Calculate annual Cash Flow as NOI minus Debt Service. NOI is provided as an input for each year. Debt Service is constant at $1,587,420. Do not introduce additional adjustments." This prevents the model from inventing reserves, CapEx charges, or other complexity that obscures the core logic.
Return Metrics
The return metrics block converts the cash flow stream into the decision variables that sponsors use to evaluate the deal: IRR, equity multiple, and average cash-on-cash return. These metrics aggregate entry cost, annual distributions, and exit proceeds into a single performance summary. Without this section, the acquisition model is incomplete—you've calculated cash flows but not returns.
Start with the Equity Multiple. This is the simplest metric: total cash returned to equity divided by equity invested. For Project Riverside, equity invested is $13,050,000 (from the Sources & Uses block). Total cash returned is the sum of annual distributions ($32,580 + $412,580 + $652,580 + $702,580 + $752,580 = $2,552,900) plus net exit proceeds ($22,440,000), totaling $24,992,900. Equity multiple is 1.91x ($24,992,900 ÷ $13,050,000). This tells the investor they get $1.91 back for every $1.00 invested, but it doesn't account for timing.
That's where IRR comes in. Internal Rate of Return is the discount rate that sets the Net Present Value of all cash flows (including the initial equity investment as a negative cash flow) to zero. In Excel, the formula is =IRR(cash_flow_range). For Project Riverside, the cash flow array is:
- Year 0: −$13,050,000 (equity invested)
- Year 1: $32,580
- Year 2: $412,580
- Year 3: $652,580
- Year 4: $702,580
- Year 5: $23,192,580 (annual cash flow + exit proceeds)
The IRR for this array is 14.2%. This is the metric that determines whether the deal meets the fund's return threshold (typically 15-20% for value-add multifamily). If IRR is below target, the sponsor either renegotiates purchase price, reduces CapEx, or passes on the deal.
Average Cash-on-Cash Return is a supplemental metric that measures annual yield on equity. It's calculated as (Total Annual Distributions ÷ Hold Period) ÷ Equity Invested. For Project Riverside, average annual distribution is $510,580, and equity invested is $13,050,000, yielding a 3.9% cash-on-cash return. This is low because most of the value creation comes from forced appreciation at exit, not operating income—a typical profile for value-add deals.
When prompting AI to generate this section, specify the exact formulas: "Calculate Equity Multiple as (sum of annual cash flows + exit proceeds) divided by equity invested. Calculate IRR using the Excel IRR function applied to the array [−13050000, 32580, 412580, 652580, 702580, 23192580]. Calculate average Cash-on-Cash as (average annual cash flow ÷ equity invested)." This level of specification prevents AI from inventing alternative definitions or misapplying formulas.
A common error is confusing project-level IRR with equity IRR. Project IRR includes debt in the cash flow calculation, while equity IRR isolates the return to equity investors. For acquisition models, always calculate equity IRR—this is what LPs care about. If you prompt "calculate IRR" without specifying equity IRR, AI may return the wrong metric.
Another mistake is calculating IRR on operating cash flow alone, excluding the exit proceeds. IRR must include all cash flows from entry through exit. If you calculate IRR on a 5-year operating stream without adding the $22,440,000 exit distribution, the result is meaningless. When building acquisition model ai, always verify that the IRR formula references a complete cash flow array including the negative equity investment at Year 0 and the final exit proceeds at disposition.
Reviewing Acquisition Logic
After generating the model, you must verify that the logic is consistent, the formulas reference the correct cells, and the return metrics match hand calculations. This is the Verification step—the final gate before presenting the model to an investment committee. Most AI-generated models contain at least one logic error, either from misinterpreting the prompt or from hallucinating cell references. Your job is to catch these before they propagate.
Start with the balance test: Sources = Uses. In Project Riverside, Sources are $22,698,000 (debt) + $13,050,000 (equity) = $35,748,000. Wait—that doesn't match Uses of $34,920,000. This reveals a loan sizing error. The correct loan amount at 65% LTV should be $22,698,000, which would make Sources $35,748,000. But Uses are only $34,920,000. This means the LTV calculation is wrong. The loan should be 65% of $34,920,000 = $22,698,000. Equity should be $12,222,000. This is the kind of error that AI introduces when you don't specify the exact sizing constraint.
Next, verify the debt service calculation. Loan amount is $22,698,000, interest rate is 5.5%, and amortization is 30 years. Using the PMT function in Excel: =PMT(5.5%/12, 30*12, -22698000), the monthly payment is $128,935. Annual debt service is $1,547,220 ($128,935 × 12). If the model shows a different number, the PMT formula is incorrect—likely due to wrong cell references or missing the negative sign on the loan principal.
Then check the exit calculation. Year 5 NOI is $2,340,000. Exit cap rate is 5.25%. Gross sale price is $2,340,000 ÷ 0.0525 = $44,571,429. Selling costs at 2% are $891,429. Loan balance at Year 5 (calculated using PPMT and IPMT functions) is approximately $21,240,000. Net proceeds to equity are $44,571,429 − $891,429 − $21,240,000 = $22,440,000. If the model returns a different exit proceeds number, either the cap rate is applied incorrectly, selling costs are omitted, or the loan balance is wrong.
Finally, confirm the IRR calculation references the correct cash flow array. The array should be seven values: initial equity investment (negative), five years of annual cash flow, and exit proceeds added to Year 5 cash flow. If IRR is calculated on only the operating cash flows (excluding exit), the result will be absurdly low. If it's calculated on exit proceeds alone (excluding operating cash flows), the result will be overstated. The correct IRR formula is =IRR({-13050000; 32580; 412580; 652580; 702580; 23192580}).
This verification process is where Decomposition pays off. Because each block is isolated, you can test them independently. If the Sources & Uses block is wrong, you don't need to check the exit calculation—fix the entry logic first, then propagate the corrected equity amount downstream. If the exit proceeds are wrong, you can verify the cap rate, selling costs, and loan balance separately, without involving the operating assumptions.
When working with AI, run these tests after every model generation. Do not assume the output is correct. AI is a drafting tool, not a verification tool. It will produce syntactically correct formulas that reference the wrong cells or apply the wrong logic. Your job is to catch these errors by running the four verification tests: balance (Sources = Uses), debt service (matches PMT calculation), exit proceeds (matches cap rate formula), and IRR (references complete cash flow array).
One final test: does the model pass the "reasonableness" check? For a value-add multifamily deal in Austin, a 14% IRR and 1.9x equity multiple is plausible. If the model returns a 40% IRR, something is broken—likely the exit cap rate is too low or the loan is sized incorrectly. If IRR is 6%, either the purchase price is too high or the operating assumptions are too conservative. Use market benchmarks to sanity-check the output. If the numbers are wildly off, don't try to debug the model—regenerate it with a more specific prompt.
This is the discipline required to build acquisition model ai reliably. AI generates the draft. You verify the logic. The model is not complete until it passes all verification tests. Most analysts skip this step and present broken models to investment committees, which destroys credibility. Do not be that analyst. Run the tests. Fix the errors. Deliver a model that works.
