Time to Hire Measures Less Than You Think

Time to hire is the most quoted number in recruiting and one of the least useful in the form most teams report it. A single average across every open role, every function and every seniority band tells you almost nothing you can act on. It goes up, someone asks why, and the honest answer is usually that the mix changed.

The metric is not worthless. It is under-specified. Decomposed properly it becomes one of the few operational signals that predicts whether a hiring plan will land.

Three definitions, three different numbers

Before anything else, agree on the clock. We have audited client reporting where three teams inside one company used three different start events and compared their numbers in the same board meeting.

• Time to fill: requisition approved to offer accepted. Includes intake delay and approval chains.
• Time to hire: candidate enters the pipeline to offer accepted. Excludes everything before sourcing starts.
• Time to productivity: offer accepted to first independent contribution. Rarely measured, most predictive of business impact.
• Time in stage: the interval that actually tells you where the process broke.

Time to fill and time to hire differ by the intake gap, and in enterprise environments that gap is often 9 to 14 days on its own. A recruiting team optimizing time to hire can look excellent while the business waits five weeks for a role to be approved and scoped.

The average hides the shape

Hiring cycle times are right-skewed. A handful of hard searches drag the tail out while the bulk of roles close in a predictable band. Reporting the mean lets one 140-day principal engineer search move the whole number and mask twenty clean 22-day fills.

Report the median and the 90th percentile together. The median tells you what normal looks like. The gap between median and p90 tells you how much variance your stakeholders are absorbing, and variance is what destroys headcount planning. A team with a 30-day median and a 45-day p90 is running a predictable operation. A team with a 30-day median and a 110-day p90 is running two operations and only one of them works.

Stakeholders do not plan around your average. They plan around your worst plausible case, and then they add a buffer you never see.

Survivorship bias in the denominator

Time to hire is computed only on roles that closed. Every requisition that has been open for 200 days and is still open contributes nothing to the number. This is the single most common way the metric lies to an operations leader.

A quarter where the hardest six searches all stayed open will show an improving time to hire, because only the easy roles completed. We recommend pairing the metric with open requisition age, reported as the median days-open of the currently active pipeline. When time to hire drops and open req age climbs in the same period, nothing improved. The hard work was deferred.

Where the days actually go

When we instrument a client pipeline, the delay almost never sits where people assume. Sourcing gets blamed because it is the visible front of the funnel. The instrumentation usually points elsewhere.

• Interview scheduling: 4 to 9 days, mostly panel coordination across time zones
• Hiring manager feedback after onsite: 3 to 6 days, the single largest recoverable block
• Offer approval and compensation review: 2 to 8 days depending on band exceptions
• Sourcing to first screen: 2 to 5 days when the channel mix is right
• Candidate decision window: 3 to 7 days and largely not compressible

Two of those five are internal coordination problems with no candidate involved. That is where the fastest wins live. Cutting hiring manager feedback latency from five days to one is a policy change, not a technology project, and it moves total cycle time more than any sourcing tool will.

Segment or do not bother

A single company-wide number is an aggregate over incompatible processes. Split it at minimum by function, by seniority, and by whether the role is a replacement or net-new growth. Replacement hires run faster because the scorecard already exists and the manager knows what good looks like. Net-new roles carry a definition cost that shows up as sourcing delay but is really a scoping delay.

Geography matters too. Our Madrid and Manila desks work against different notice period norms, and a 30-day statutory notice period is not a performance problem. Reporting it inside the same average as an Austin two-week start is how a good team ends up defending a number that was never in its control.

What we report instead

For Novexhire engagements we report four numbers per requisition family: median time in each stage, p90 total cycle, open requisition age, and offer acceptance rate. The fourth one keeps the first three honest. It is trivially easy to compress cycle time by lowering the bar, and acceptance rate plus 90-day retention is what catches that.

None of this requires new tooling. It requires deciding what event starts the clock, refusing to report a mean without a p90 beside it, and publishing open requisition age in the same table. Teams that do those three things stop arguing about whether hiring is slow and start arguing about which stage to fix, which is a much more productive argument.