Four Numbers That Tell You Whether Your Technical Hiring Works

Most hiring dashboards are activity reports. Interviews held this month. Candidates sourced. Applications received. Pipeline by stage, all of it going up and to the right, none of it answering the only question that matters: is this process producing good hires, and at what cost to the people who go through it?

Four numbers do answer that. Three of them are uncomfortable, which is the main reason they are rarely tracked.

One: how long a candidate waits between stages

Not time to hire. The gaps inside it.

Time to hire is an aggregate that hides what candidates experience. What they feel is the eleven days between the screen and the technical interview, and the six days of silence after the final round while an internal decision gets made.

How to measure it. Export stage transition timestamps from your applicant tracking system for the last twenty candidates and look at the median gap per stage, not the total. Twenty is enough to see the shape.

What it tells you. Where you lose people. The longest gap is almost always an internal scheduling problem rather than a candidate problem, which means it is yours to fix and usually cheap. Reduce time to hire for engineers goes through where the days actually go.

Two: what share of people who start your assessment finish it

This is the cleanest signal you have about whether the thing you are asking is reasonable.

If a quarter of candidates start your take-home and never submit, that is not a commitment problem. That is a scope problem, a clarity problem, or an environment problem, and every single one of those is yours.

How to measure it. Count starts and submissions over a quarter. If your assessment runs in a hosted environment, you already have this. If it is a repository you email out, approximate it by invitations versus returns.

What it tells you. Whether your assessment is sized for the market you are hiring in. It also moves immediately when you fix the environment: a large share of non-completions are people who lost an evening to setup and gave up. Why your best candidates drop out of the funnel is the long version.

Three: how often two interviewers reach the same verdict

This is the number almost nobody measures, and it is usually the reason the loop keeps growing a stage at a time.

When interviewers disagree frequently, teams do not fix the calibration. They add a round. Then another. Five stages later, the process is long, candidates drop out, and the disagreement is still there, now spread across more people.

How to measure it. For any stage where two people score independently, compare their verdicts: simple agreement on hire versus no hire is enough to start. You do not need a statistic, you need a percentage and a trend.

What it tells you. Whether your scorecard means the same thing to different people. If agreement is low, the fix is a shared rubric and calibration sessions, not another interview. The bar raiser problem is about exactly this.

Four: how your hires are doing at six months

The only outcome measure in the list, and the hardest to collect honestly.

How to measure it. Not a performance score. Ask the hiring manager two questions at six months: would you hire this person again, and what surprised you. Record both. After ten hires you have a pattern, and after twenty you can compare it against what the interview predicted.

What it tells you. Whether your process predicts anything at all. The surprises are the valuable half: if three managers say the same thing surprised them, your interview has a blind spot with a name, and you can add one task that covers it.

What to ignore

Number of interviews conducted. Activity, not outcome.

Offer acceptance rate on its own. It moves with compensation and market conditions far more than with process quality, so it tells you about your salary bands.

Any metric that nobody will act on. A number that cannot change a decision is a chart that costs a meeting every month.

How to start without a project

Pick one, this quarter, and measure it for twenty candidates. The waiting time is the easiest and usually the most actionable, so start there unless you already know it is fine.

Then make one change and measure again. That is the entire method. A hiring process improves the way any other system does, by changing one thing and watching a number, and it is remarkable how rarely anyone does that.

Where EasyEnv is different

Two of these four numbers are hard to collect if your assessment is a repository you email out: you cannot see who started, who got stuck on setup, or what happened before the submission.

When the assessment runs in a hosted environment, starts, completions and abandonment are observable by default, and the session recording tells you where somebody gave up, which is the difference between knowing the completion rate and knowing why. Because every candidate works from the same recipe, comparing two people, or two months, is comparing like with like.

The limits, plainly. The six month number has no tooling answer: somebody has to ask two questions and write the answers down. Interviewer agreement needs two independent scores, which some teams will not organise. And none of these numbers is meaningful on a sample of five hires a year, which is most small companies: there, read the surprises rather than the percentages.

The short version

Stop reporting interviews held. Measure the gap between stages, the share of candidates who finish what you ask, how often two interviewers agree, and what hiring managers say at six months.

Pick one this quarter, measure twenty candidates, change one thing, measure again. The waiting time is the cheapest to fix and the one candidates feel most.