Hiring Analytics

Where candidates leave

Count how many candidates enter and leave each stage, and how many of the departures were your decision rather than theirs. That second distinction is the one most dashboards omit and the one that carries the information.

Candidates withdrawing in numbers points at your process: it is too slow, too long, or something in it is putting people off. Candidates being rejected in numbers at one stage points at the stage before it, which is admitting people it should not.

Both are actionable. Total applications, the number most reported, is not.

Time, measured usefully

Total time to hire is a summary, and a summary cannot be fixed. Break it into the gaps between events: posted to first review, review to first interview, interview to decision, decision to offer, offer to acceptance.

Those gaps are where the weeks live, and each has a different cause and a different fix. The gap between final interview and offer deserves separate attention because it is entirely internal and lands when the candidate is most likely to have another offer arriving. See Reducing Hiring Time.

Source quality

Track sources by hires, and where you can by hires still there a year later. Applications per source is close to a vanity measure: a channel that sends four hundred applications and no hires has cost you review time and produced nothing.

This is one of the few places where a small amount of data changes behaviour quickly, because the differences between channels are usually large and obvious once counted at all.

Check your own criteria

The highest-value analysis in hiring is also the least performed. After a handful of hires, look back at the screening criteria and ask which of them actually predicted the people you kept.

The usual findings are consistent: one or two requirements did nothing, one was excluding capable people, and something not on the list at all turned out to matter. Because scoring keeps a record, that review is a reading exercise rather than a research project.

The same exercise applies to interviews. If two interviewers consistently disagree, one of them is measuring something the rubric does not describe, and that is worth knowing either way.