The metrics most engineering recruiting teams track are the wrong metrics. Not because they are inaccurate, but because they measure the wrong things. Resume views, inbound application volume, funnel conversion rates, time-to-fill: these are activity metrics. They tell you how busy your recruiting process is, not whether it is working. The difference matters more than most teams realize when they are in the middle of a hiring cycle.
There is a category distinction that helps clarify this. Leading indicators predict future outcomes; you can act on them early. Lagging indicators confirm past outcomes; by the time they tell you something is wrong, the hiring cycle they describe is already over. Most recruiting metrics are lagging. And among lagging indicators, the most commonly tracked ones are not even measuring the outcome you care about most.
The time-to-fill problem
Time-to-fill is the most reported recruiting metric. It measures how many days elapsed between a role opening and an offer acceptance. It is easy to calculate, easy to understand, and it tells you almost nothing useful about whether you hired the right person.
A team that rushes through screening and extends an offer quickly will have a low time-to-fill. A team that moved methodically and made a better decision will have a higher one. Neither number tells you whether the person who was hired was the right person for the role. And the outcome that actually matters, which is whether the hire worked out, is typically not measured at all or is measured so infrequently (annual reviews, six-month check-ins) that the connection between the hiring process and the outcome is nearly invisible.
Optimizing for time-to-fill also creates specific failure modes. Teams under pressure to reduce time-to-fill will compress evaluation stages, which increases the probability of a poor fit hire. They will also be more likely to extend offers to candidates who are adequate rather than waiting for candidates who are strong, because waiting increases time-to-fill. These are rational responses to the wrong incentive.
Offer acceptance rate as a vanity metric
Offer acceptance rate is widely reported as a marker of employer brand and compensation competitiveness. In some contexts it is useful. But as a primary metric, it has a fundamental problem: it only measures the end of the funnel, and what happens at the end of the funnel is heavily shaped by everything that happened before it.
A team with a 95% offer acceptance rate might be very good at recruiting. Or they might be extending offers very selectively, only to candidates who have already signaled strong intent, and declining to extend offers to candidates who might say no. That would also produce a 95% acceptance rate while obscuring a lot of lost candidates earlier in the funnel.
The more useful version of this question is not "what percentage of our offers are accepted?" but "of the candidates who accepted our offer, how many were our top-ranked candidate at offer stage?" If you are consistently making an offer to your third-choice candidate because your first and second choices declined or got to offer stage at another company first, your acceptance rate may look healthy while your actual hiring quality is degraded.
What to measure instead
The metrics worth tracking are the ones that predict hiring quality and allow you to intervene before the hire is made.
Screen-to-interview conversion rate by source. What fraction of candidates from each sourcing channel pass the initial screen? If referrals pass at 40% and job board applicants pass at 8%, that tells you something about the quality differential between channels. Tracking this over time tells you whether a sourcing change improved quality before you have to wait for hire outcomes to confirm it.
Interview-to-offer rate. What fraction of candidates who enter the technical interview stage receive an offer? If this rate is very high (above 50%), you may be letting too many candidates through the screen who are not actually qualified for the role. If it is very low (below 10%), the screening criteria may not be well-calibrated to the interview criteria, or the interview is evaluating for something different than the role requires. Both extremes suggest a calibration problem.
Interviewer prediction accuracy. After a hire is made, go back and compare the assessment scores from interviewers to the six-month performance outcomes. Which interviewers were consistently accurate in predicting performance? Which were not? This is harder to instrument but produces the highest-quality insight because it tells you specifically where your evaluation process has signal and where it has noise.
First-year retention by hiring source. This is a lagging indicator, but it is measuring the right thing. A hire that leaves in nine months due to poor fit represents a process failure regardless of how good the time-to-fill and acceptance rate looked. Tracking this by channel and by hiring manager over time builds a picture of which processes and which interviewers produce durable hires.
The calibration problem under the metric problem
The deeper reason recruiting teams track activity metrics instead of quality metrics is that quality metrics require calibration that most teams have never done. To know whether your interviewers are accurate, you have to define what accurate means, which requires defining what success in the role looks like, which requires having had that conversation before you started hiring.
Most teams have not had that conversation explicitly. They have a shared intuition about what a good hire looks like, but that intuition is not written down and is not consistent across everyone involved in the hiring process. When the hire does not work out, the post-mortem is often framed as "we got unlucky" or "the candidate misrepresented themselves" rather than "our evaluation criteria did not predict what we needed."
This is the fundamental signal-to-noise problem in engineering recruiting. The noise is not just the volume of resumes or the imprecision of keyword matching. It is the lack of a clear signal to optimize toward. If you do not know what you are trying to measure, no metric will tell you whether you are measuring it well.
Starting from outcomes and working backward
The path to better recruiting metrics starts from outcomes, not from the process. What does a successful hire look like at six months in this specific role? What technical capabilities do they need, and what working style does the team require? Write those down. Then design your evaluation to test for exactly those things. Then measure whether your evaluation correctly predicts the outcome.
That loop, outcome definition to evaluation design to prediction accuracy measurement, is what turns a recruiting process from an activity-tracked workflow into a signal-generating machine. It takes more upfront work per role. It produces meaningfully better decisions per hire. And over time it builds the institutional knowledge about what actually predicts success on your specific team that no tool or benchmark can substitute for.
We built Fonzi around this idea: that the match dimensions worth measuring are the ones that predict role success, not the ones that are easy to count. Getting that right at the shortlisting stage makes every downstream step more efficient, because you are working with candidates who are genuinely worth evaluating deeply rather than a volume pool optimized for surface metrics.
