Cap Rate vs Cash-on-Cash vs IRR: When Each One Lies

Four constructed deals where cap rate, cash-on-cash and IRR each give the wrong answer. Full arithmetic at 6.67% money, plus which metric to lead with.

Cap rate measures the asset, cash-on-cash measures your capital in year one, and IRR measures every dollar across the whole hold including the sale — so all three can be correct and still point three different directions on the same property. The useful question is not which metric is best. It is which one is lying to you on the deal in front of you.

Most articles define the three formulas and stop. This one builds four deals where the metrics openly disagree, shows all the arithmetic, and names each one's failure mode.

Cap rate describes the building, and deliberately ignores you

Cap rate is NOI divided by purchase price. NOI is effective gross rent minus operating expenses and the CapEx reserve — no mortgage payment, ever. That omission is the point: cap rate compares buildings on equal footing regardless of who buys them or how, so a cash buyer and a 75%-leveraged buyer compute the identical number on the same property. Which also means it cannot see your loan, your rate, your rehab budget or your closing costs. In most markets a 6–8% cap rate is strong; below that you are buying growth, not yield.

Cash-on-cash measures year one, and only year one

Cash-on-cash is annual cash flow divided by total cash invested — down payment plus rehab plus closing costs plus points. It is the most honest metric about your position, because the denominator is real money that left your account and the numerator is real money arriving. Its blindness is time. It has no opinion about principal reduction, rent growth, or what the property is worth when you sell. A deal that returns 5% forever and a deal that returns 5% in year one on its way to 11% look identical.

IRR measures every dollar including the exit, which is its strength and its weakness

IRR is the discount rate that sets the net present value of the full cash flow series to zero — the initial outlay, every year of cash flow, and the net sale proceeds at the end. CasaWise solves it with Newton-Raphson over up to 100 iterations (how the model is built). It is the only one of the three that prices time, amortization and the sale into a single number. And the sale is a forecast. IRR inherits every appreciation assumption you made, with compounding — the most complete metric and the easiest one to manufacture.

Four deals, every input on the table

These are constructed, not observed — each pair isolates one variable so you can watch a metric fail. All four use a 30-year fixed at 6.67% (FRED, MORTGAGE30US, week of 13 Aug 2026), 8% vacancy, a five-year hold, and the standing 6% selling-cost assumption. Operating cash flow is held flat so the exit is the only thing moving between deals.

Deal A's expense stack, itemized, so the $9,350 is real lines and not a plug:

Property tax        1.10% of price      $2,750
Insurance           0.45% of price      $1,125
Maintenance/repairs 5% of gross rent    $1,500
Property management                     $2,175
CapEx reserve       6% of gross rent    $1,800
------------------------------------------------
Total operating expenses                $9,350

Effective gross rent   $2,500 x 12 x 0.92 = $27,600
NOI                    $27,600 - $9,350   = $18,250

The results, with the exit computed as Property Value x (1 − 6%) − Remaining Loan Balance:

Deal A vs Deal B: identical 7.30% cap rates, and a 37% gap in cash-on-cash

Both yield exactly 7.30% unlevered, so a cap-rate-first buyer treats them as interchangeable and picks on location, roof age, tenant file. Then the financing lands.

Deal A   $18,250 cash flow / $255,000 invested = 7.16% cash-on-cash
Deal B   $6,042 cash flow  / $133,000 invested = 4.54% cash-on-cash

Two things cap rate could not see. First, Deal B's $25,000 of deferred maintenance and $8,000 of closing costs sit in the denominator of cash-on-cash and nowhere in cap rate — cap rate divides by purchase price, not by what the deal costs you.

Second, and larger: at 6.67% over 30 years the annual loan constant is 7.72% — $643.29 a month per $100,000 borrowed, times twelve. When the constant exceeds the cap rate, every borrowed dollar earns 7.30% and costs 7.72%. Leverage runs backwards. Buy Deal B for all cash and its cash-on-cash is $29,200 / $433,000 = 6.74%; the loan removes 2.2 points.

That is not an edge case in 2026, it is the default condition. A property has to clear roughly a 7.72% cap before borrowing at 6.67% improves your yield, and most do not. Cap rate lies whenever you compare your deal to someone else's, because it was never measuring your deal.

Deal B vs Deal C: identical 4.54% cash-on-cash, IRR of 7.34% against 3.07%

Now hold cash-on-cash constant. Deal B returns 4.54% on $133,000; Deal C returns 4.54% on $306,000. In year one they are the same investment at different scale, and a cash-on-cash-first buyer takes Deal C — more absolute cash flow, no lender, no vacancy risk against a mortgage payment. Over five years they are nothing alike.

Deal C throws off more than twice the cash and finishes worse, because two of the four pillars never show up. An all-cash purchase has no loan to amortize. Appreciation was assumed at zero. And the 6% selling cost still arrives — $18,000 handed back at the closing table against a value that did not move. Meanwhile Deal B's $300,000 loan amortizes $18,624 while 3% appreciation adds $63,710 to a $400,000 basis. Both are invisible to cash-on-cash. Cash-on-cash lies whenever the hold period matters, which is every buy-and-hold.

Deal C vs Deal D: the best IRR in the set belongs to the worst deal

Deal D posts the highest IRR of the four — 11.02%, against 8.32%, 7.34% and 3.07%. It also has the lowest cap rate, a 0.58% rent ratio, and it bleeds $7,948 a year. Here is where the 11.02% comes from:

Operating cash flow, 5 years   -$7,948 x 5      = -$39,740
Net sale proceeds              $669,113 x 0.94
                               - $351,720       = $277,246
------------------------------------------------------------
Total profit on $135,000 invested               = $102,506

Total cash coming back over five years is $237,506, and the sale supplies $277,246 of it — 117% — because operations remove $39,740 on the way. The sale is not the largest contributor to the return; it is the only one. Every dollar of that $102,506 is produced at a single moment, by a single assumption, five years out.

Test the assumption:

Drop appreciation from 6% to the 3.0% CasaWise uses as its base default and the IRR falls from 11.02% to 2.33%. Nothing about the building changed — no tenant left, no roof failed, no rate moved. One assumption moved three points and took 79% of the return with it.

Watch the 6% selling cost across those rows. At 6% appreciation it takes $40,147; at 0% it takes $30,000 against a gain of nothing, which is why Deal D at flat values does not merely underperform — it loses $56,460. Exit cost is the one line in an IRR that is certain while everything above it is a forecast.

IRR lies whenever the terminal value carries the number. An IRR built on appreciation is a forecast wearing a metric's clothes. Before you accept one, do what that table does: re-run it at the base rate, then at zero, and see what survives. More in appreciation and rent growth assumptions.

Why CasaWise weights IRR at 40% and still refuses to print it sometimes

IRR carries the largest single weight in the Deal Score — 40%, on a curve where 0% is the floor, 9% the midpoint and 18% the ceiling. It earns that weight because it is the only metric that prices the full hold. It does not get all of the weight for exactly the reason Deal D demonstrates. Deal D posts the best IRR in the set and scores 28.6 — Caution, because the other 60% is doing its job: cash-on-cash at 25% scores zero at −5.89%, DSCR at 20% scores zero at 0.73, and a 4.20% cap rate and 0.58% rent ratio add almost nothing. The metric with the biggest weight said buy. The system said no. That disagreement is the feature.

Two guardrails sit on top. IRR is displayed only within −100% to +500% — outside that band the figure is arithmetically valid and practically meaningless, usually the artifact of a near-zero denominator or a series that flips sign more than once. When the solver does not converge inside 100 iterations, CasaWise reports N/A rather than the last iterate. A confidently wrong IRR is worse than none, because you will act on it.

The metrics nobody weights properly

DSCR is a veto, not a score. NOI divided by annual debt service, and it belongs to the lender before it belongs to you. Deal B is 1.26 and financeable. Deal D is 0.73 — the property covers 73 cents of every dollar of debt service before a single maintenance surprise. Most lenders require 1.20+, and below 1.00 the outcome is not a weak return; it is monthly contributions from outside income, with foreclosure at the end of a bad enough run. DSCR carries 20% of the Deal Score, floor 0.95, ceiling 1.40. No IRR fixes a 0.73.

GRM and rent ratio are screens, nothing more. Deals A and B share an identical 8.33 GRM and an identical 1.00% rent ratio, and their cash-on-cash returns differ by 37%. GRM ignores expenses entirely; under 12 warrants a closer look, and that is the whole of its usefulness. The rent ratio is the 1% rule with the dogma removed — CasaWise grades it on a curve where 0.65% is par and approaching 1% is excellent, weighted at 5%. Use both to eliminate addresses in seconds, never to choose between survivors.

One tension worth naming: professional investors have historically targeted 8–12%+ cash-on-cash, while the Deal Score curve tops out at 6% because that is what is achievable against 6.67% money. Deal A's 7.16% scores 100 today and would have been mediocre in 2015. Both are true — calibrate to the rates you are actually buying in. See what a good rental deal actually looks like.

Which metric to lead with

• Buy-and-hold. Lead with cash-on-cash, confirm with DSCR, use IRR to break ties between deals that already cash flow. You cannot eat appreciation; cash flow is the only liquid pillar.
• Value-add. Lead with IRR — the thesis is a change in NOI over time that cash-on-cash cannot represent. Then check year-one DSCR against your worst case; the lender underwrites the property you bought, not the one you intend to create.
• Flip. None of the three. Use flip ROI and hold months — cap rate on a seven-month hold is noise, and IRR on a sub-year hold annualizes into an absurd figure. Holding costs decide flips.
• Refinance / BRRRR. Lead with post-refi DSCR and post-refi cash-on-cash on your remaining capital after cash recouped. The refinance resets the denominator, and any metric measured against the original basis stops describing your position the day it funds.

Run every property through all of them. The most useful signal in underwriting is not a strong number — it is two metrics disagreeing, because that always has a specific cause, and the cause is usually what decides the deal. How to analyze a rental property walks one deal end to end.

Frequently Asked Questions