The $17.7M Cost of a Single Hiring Filter

At one Fortune 500 insurance carrier, a single screening rule, "must have insurance experience," would have rejected 2,863 agents who went on to produce. Those agents represented $17.7M in observed annual production. The retrospective result does not establish causality or realized savings. It measures production that would have been at risk if the rule had been enforced across the study cohort of 10,765 agents hired.

Source: "Decision Traces," Saad Bin Shafiq, Nodes, 2026. N=10,765 agents hired 2022 to 2025 at a Fortune 500 insurance carrier, analyzed inside the carrier's VPC. Read it on arXiv.

Estimate a filter's cost on your own numbers: Hiring ROI Calculator.

How one filter loses $17.7M

The "insurance experience required" rule looks reasonable. At this carrier, agents with insurance experience produced at 28.0%, while agents without it produced at 33.7% (odds ratio 0.763, anti-predictive within multiple hiring years). Of the agents who lacked the flag, 2,863 went on to produce. Their observed annual production totaled $17.7M. Enforcing the rule as a hard gate would have excluded them from this funnel.

The candidates the filter would have cut

Two populations tell the whole story:

No keywords, high score: 677 agents had no traditional ATS keywords and fell in a high behavioral-score band. They produced at 33.7%, above the overall rate.
Every keyword present: agents who matched all the standard filters produced at only 26.3%, below the no-keyword high-scored group.

The signal the carrier trusted was inversely related to the outcome it wanted.

Why this stays hidden

The ATS records who you screened out. The HRIS records who produced. The rule and its outcome had been stored in separate systems, so the cost of the filter did not appear in either system alone.

What each system can see

System	What it knows	What it cannot see
ATS	who you screened and on what	whether they produced
HRIS	who produced and how much	what you screened them on
Decision trace	both, connected	outcomes or fields the source systems never recorded

Find your own number

The $17.7M is specific to this carrier and this role. Another organization can test its own filters by connecting ATS screening rules to defined HRIS outcomes, then measuring the observed outcomes of people each rule would reject. The result depends on that organization's data, role, cohort, and outcome definition.

Frequently asked questions

How much can one bad hiring filter cost? In this study, requiring insurance experience would have rejected 2,863 agents who produced, representing $17.7M in annual premium credit.

How was the $17.7M calculated? It is the sum of observed annual premium credit across the 2,863 producing agents who lacked the insurance-experience flag and would have been excluded by keyword-based filtering.

Do candidates with no keywords perform worse? Not here. Agents with no traditional keywords in a high behavioral-score band produced at 33.7%, above the overall rate, while agents with every keyword present produced at 26.3%.

How would we find our own version of this number? Connect your ATS screening rules to HRIS outcomes through a decision trace, then add up the production of the people each rule would screen out.