Time to Hire

Plans slipping because a role went unfilled is common, and why it slipped rarely gets explained. Hiring is work several people touch a little, and each of them believes they didn't cause a delay. The person reviewing resumes was only three days late. The interviewer was only a week late finding a slot.

Small delays like that stack into an entire quarter. At Pagedeck, Theo's request for one engineer was four months old, and when Grace asked why, nobody could answer. Reconstructing the record found three weeks waiting on resume review, two weeks finding interview slots, and two weeks waiting on approval of the offer amount. The time actually spent interviewing, across those four months, was two days.

That's what this number is for. It isn't a search for who is slow — it makes visible the stretches when nobody was working on it at all. It counts waiting, not people, which is also why it reads as uncomfortable the first time a team looks at it: most of the elapsed time belongs to nobody.

Subtract the start point from the end point. Five steps.

1. Fix the start and end points and write them down. The start is the day the job description was posted, or the day the hiring request was received. The end is the day the offer was accepted, or the day the person actually started. Either convention works, but companies differ, so check which one anyone else's figure used before comparing. Request-received to start-date runs roughly six weeks longer than posted-to-accepted.
2. Count only completed hires. Requests cancelled midway or roles that disappeared come out of this number and get counted separately. Mixed in, the ones that dragged on before being abandoned inflate the figure.
3. Use the median, not the mean. Five hires at 28, 31, 35, 40, and 120 days average to 51 days and have a median of 35. The point is to stop one unusual search from dragging the whole number; the 120-day case gets pulled out and examined on its own.
4. Break it down by step. This figure alone points at nothing to fix. Count days to resume review, to first interview, to final interview, to offer, and to start date. Pagedeck's engineering hires last quarter came to 12, 9, 14, 11, and 30 days.
5. Split it by role and level. One company-wide number means whichever role hires most becomes the number.

Step 4 is what separates fixable from not. The 12 and the 14 are stretches where nobody was working, so they can shrink. The final 30 is the notice period at the hire's current employer, which usually can't. Without separating the controllable boxes from the uncontrollable ones, target-setting produces numbers nobody can meet. A target of 30 days end to end, in a market where notice periods run a month, is an instruction to make offers to people who are already unemployed.

There is no absolute standard, and roles differ by more than a factor of two. A role with many applicants closes in three to five weeks; a senior engineer with narrow requirements at eight to twelve is ordinary. The more a role depends on sourcing — going out to find candidates rather than waiting for applications — the longer the early stretch runs.

Shorter isn't automatically better, either. Lower the bar and this figure drops immediately, while the people hired take longer to become productive and more of them leave inside six months. So the number never travels alone. Read it against offer acceptance rate, the share of offers accepted, and against six-month retention. If the duration fell while acceptance fell with it, offers were going out in a hurry to whoever was available.

Compare against your own last quarter rather than an industry average. Company size, employer brand, and location all differ, which makes somebody else's number hard to use. Filling the same role in 47 days last quarter and 61 this quarter is a question worth asking, and the step-level breakdown usually answers it in one line.

The lever that actually moves in practice is waiting time per step. A rule that resume review finishes within 48 hours, a rule that interviewers keep two slots open each week, a decision to push offer approval down to the hiring manager — each takes days off. Cutting interview rounds also shortens the duration, but it removes evidence for the decision, so look at waiting time before touching the number of steps.

"Time to hire is 61 days and 26 of them are waiting. Keep the interview rounds and fix how we schedule."

The problem is the waiting stretches, not the number of steps.

"That figure is from request received. Measured from posting it's six weeks shorter."

Different conventions make the comparison invalid.

"Duration dropped and offer acceptance dropped with it. More offers are going out rushed."

Naming the result of managing on speed alone.

After reconstructing Theo's four months, Pagedeck set three rules. Resume review finishes within 48 hours of receipt, interviewers keep two slots open each week, and offers inside the pay band get finalized by the manager.

The next quarter's median for engineering hires went from 61 days to 43. All 18 days came out of waiting stretches, with interview rounds and the hiring bar unchanged. Offer acceptance over the same period rose from 62 percent to 71 percent, apparently because candidates had less time to sit weighing another company's offer.

The most common reading is that shorter is always better. Duration metrics usually exist to be reduced, so it looks that way. But set this as a target and the fastest route is lowering the bar. Choose among three applicants instead of waiting and six weeks becomes three, and the same role gets filled again six months later. Left uncorrected, hiring speed shows as improved on the chart while the team stays short-staffed.

Treating it as the recruiting team's performance metric is the second mistake. Recruiting runs the process, so the assignment feels natural. In practice most of the delay accumulates elsewhere: the people reviewing resumes, sitting in interviews, and approving amounts each contribute days. Hang this number on recruiting alone and recruiting gets judged on something it can't control, while the people actually holding the days receive no signal at all. Splitting it by step and attaching each box to whoever owns it is what earns its keep.