Nearly every deal that fails was underwritten as a good deal. That is worth sitting with, because it means the failure was not in the arithmetic. The spreadsheets were usually correct. The inputs were plausible. The conclusion followed from the assumptions.
The failure was that the assumptions all described the same scenario — the one where things go according to plan — and no one ran the deal against any other.
What best-case underwriting looks like
It rarely looks reckless. It looks like a reasonable analysis where every individual input sits at the optimistic end of its plausible range.
The ARV is the top of the comp range rather than the middle. The renovation budget is the contractor's number without contingency. The timeline is the contractor's schedule without permit delay. Days on market is the fastest recent sale rather than the average. Rent is the highest current listing rather than signed leases. The refinance is priced at today's rate rather than a rate four months out.
No single one of those is unreasonable. Each is defensible in isolation. But the probability that all seven land at the favorable end simultaneously is small, and the analysis has silently assumed exactly that.
This is the core error. Optimism does not have to be extreme to be fatal. It only has to be consistent.
Why the errors compound rather than average
Investors intuitively expect variance to cancel out — some things run over, some run under, it evens up.
It does not, for two reasons.
The errors are correlated. A soft market simultaneously lowers your ARV, extends your days on market, increases your holding costs, and weakens the rental comps supporting your refinance exit. These are not independent draws. They are one condition expressing itself across four line items.
Margin is a residual on top of large gross numbers. A flip with a $400,000 ARV and a $340,000 all-in cost has a $60,000 margin — 15% of the gross. A 10% ARV miss is $40,000, which is two-thirds of the margin. Small percentage errors in the large numbers produce large percentage swings in the small one.
Combine both effects and modest, individually reasonable optimism across several inputs reliably produces a loss.
The stress test
The correction is not pessimism. It is running the deal at more than one scenario and making the decision with all of them visible.
Take your base case and adjust each of these independently, then together:
Resale value down 10%. Not a crash — normal variance between an optimistic comp read and an actual sale.
Renovation cost up 20%. Change orders, unforeseen conditions, material price movement. This is a routine outcome, not a bad one.
Timeline extended four months. Permit delay plus construction overrun plus slower absorption. Apply your real monthly holding cost, not a nominal one.
Rents down 10%. For any deal where the exit or the backup exit is a hold.
Refinance rate up 100 basis points. For any BRRRR or refinance exit.
Run each individually to see which input the deal is most sensitive to. Then run all of them at once. That combined case is not a worst case — it is a mildly unfavorable year, and it happens frequently.
Reading the result
The combined stress case will usually show a loss. That alone does not disqualify the deal. What matters is the shape of the answer.
How large is the loss, and can you absorb it? A stress case showing a $15,000 loss on a deal with $60,000 of upside is an acceptable risk profile if you have liquidity. A stress case showing a $120,000 loss on the same upside is not, regardless of how attractive the base case looks.
Which input drives the result? If the deal collapses on the ARV adjustment, the ARV needs independent verification before you proceed — that is where the risk is concentrated. If it collapses on timeline, your reserves and loan term need to accommodate a longer hold.
Does a viable alternative exist in the downside? A flip that becomes a break-even rental if the resale market softens has a floor. A flip that only works as a flip has none. The presence of a real second exit changes the risk profile more than any amount of margin does.
How much delay can it survive? Take your margin and divide by your true monthly holding cost. That number — months of runway — is a better single risk indicator than projected return percentage.
Where investors resist this
Two objections come up consistently.
"If I stress test everything, no deal will pencil." Some deals will. The ones that survive a reasonable downside are the ones worth doing. A pipeline where nothing passes a mild stress test is not a stress test that is too harsh — it is a market where the available deals do not have enough margin, and the correct response is patience or a different market, not a looser standard.
"I have done this before and it worked out." Past outcomes in a rising market do not validate the analysis. Appreciation covers a great deal of underwriting error and stops covering it without notice. An approach that worked for four years in a strong market has not been tested.
The incentive problem
There is a structural reason best-case underwriting persists, and it is not that investors lack analytical ability.
Nearly every number in a deal package arrives from someone compensated on the transaction closing. The wholesaler's ARV. The agent's comp opinion. The contractor's bid and schedule. The lender's willingness to size against a projected value. Each of those parties is competent and most are honest, but none of them is paid to tell you the deal does not work.
The only party with an interest in the downside is you, and you are also the party who wants the deal to work. That is a difficult position from which to run an objective stress test on your own analysis.
This is the case for independent review — not because investors cannot do the math, but because the person running the numbers should not be the person who wants a particular answer.
The practical standard
Before capital is committed, three documents should exist: the base case, the stress case, and a written statement of which assumptions the deal is most sensitive to.
If those three things exist and the deal still looks acceptable, proceed. If only the base case exists, the deal has not been underwritten. It has been modeled.
