What a screening override really costs, and how to measure your exception rate

An override is admitting an applicant the screening flagged. It is not automatically a mistake, and treating it as one is how you end up rejecting good applicants.

Some overrides are legitimate: a cash-paid worker the automated path could not verify, a guarantor arrangement the system did not model, a household composition the flow did not anticipate. Others are a leasing agent under pressure to fill a unit.

You cannot tell which is which from an override count. You need the outcome.

Measure it in four steps, and none requires new software

Define the population. Applicants the screening recommended against who were admitted anyway. That is your bypass population.

Define delinquency the same way every time. A workable definition: lease status current, and pending balance at or above two times monthly rent. Pick one and hold it, because a shifting definition makes the trend meaningless.

Split current delinquency by what the screening recommended. Bypass-admitted against approved. Two rates.

Divide. Bypass delinquency rate over approved delinquency rate gives you a risk multiplier, which is the single most useful number in this whole exercise.

The 3-to-6-month lag is why nobody connects the two

A bypass admission and its consequence are separated by a quarter or two.

A resident admitted in February with unverified income does not go delinquent in March. They go delinquent in June or August, by which point the approval decision is four months old, the leasing agent may have moved on, and nobody is looking back at the application.

That lag is the entire reason override rates drift upward without anyone deciding they should. The feedback loop is too slow to be felt.

Which means the only way to manage it is to measure the cohort deliberately rather than waiting to notice.

Read the override rate as a trend, not a level

Override rate for a month is bypassed admissions divided by total fails that month.

The level is hard to judge in isolation. The trend is not. A rate moving from 40% toward 12% over a couple of quarters is a real operational change. A rate holding steady at 15% while delinquency rises means something else is going on.

Read it per property too. Overrides concentrate: a property under lease-up pressure or with a hard-to-fill floor plan generates far more than a stabilized asset, and the portfolio average hides it.

Categorize the reason, or the number tells you nothing actionable

An override count is a number. An override count by category is a work list.

Four categories worth separating:

Cash or non-traditional income. The applicant genuinely earns enough and the automated path could not confirm it. This is a product gap on our side, not a bad decision on yours.

Guarantor or household arrangement. The applicant does not qualify alone and somebody else is standing behind them. Fine, if your policy says so in writing.

Document could not be verified. Not fraud, just an unreadable or stale document. Usually fixable by asking again.

Fraud flag overridden. The one that should be rare and reviewed individually.

If most of your overrides are the first category, the fix is better verification coverage for cash-paid applicants rather than tighter approval discipline. Automated employer confirmation is requested on our roadmap rather than shipped, and it is aimed exactly at this.

If most are the fourth, that is a different conversation.

What the exposure looks like when you total it

The reason to do this arithmetic is that it converts a policy debate into a number.

For each delinquent bypass tenant, sum the delinquent balance plus any eviction balance. That total is your financial exposure from override decisions over the period.

Be explicit about what it excludes: legal fees, unit turn costs, and lost rent during vacancy. So the real cost is higher than the figure, and the figure is still usually enough to end the debate.

Overrides have to be logged to be measurable

This is the mechanical precondition and it is where most portfolios fail before they start.

If a leasing agent can admit a flagged applicant without the decision being recorded against the application, none of the above is computable. You have a delinquency rate and no way to attribute it.

Every override in ProofUp is logged, with the flag it overrode. Not to police anybody. Because an override rate you cannot see is a delinquency source you cannot measure, and the measurement is what lets you defend the legitimate overrides.

The reference portfolio's outcome after making this visible: delinquency down 50%, evictions down 24%. Applicants approved without verified income default 4x more often, which is the number that makes the case on its own.

Run it for last quarter

Pull your flagged-and-admitted list and your approved list. Compute delinquency for each. Divide.

If the multiplier is above 3x you have a measurable, attributable cost and a category breakdown worth doing next.

What is your multiplier?