Between 2023 and 2025 we documented 58 placements. The median ran 43 days from kickoff to accepted offer. That number is the headline, and it is the least interesting thing on the page.
The interesting thing is what the dataset says about why some searches finished near the median and others ran long. Once you have enough of them recorded the same way, the pattern stops being anecdote. The searches that ran long did not run long for the reasons most teams brace for. They ran long for the reasons most teams ignore.
Here is the finding, stated plainly: speed is a byproduct of structure. The fast searches were not staffed by faster recruiters or luckier pipelines. They were better specified up front, and the specification paid out at every stage that followed.
Long searches stall at the decision points
The intuition is that a slow search is a sourcing problem, that the days pile up because the right people are hard to find. In our data that is mostly wrong. The pipelines on long searches were not noticeably thinner. The candidates were there.
The searches stalled at decision points. A strong candidate sat in an inbox for a week because no one owned the next step. A debrief slipped because two stakeholders could not find an hour. A maybe lingered as a maybe because the team had never agreed what a yes required. None of that is sourcing. All of it is decision latency, and it is where the weeks actually go.
Speed is a byproduct of structure. The fast searches were not staffed by faster recruiters. They were better specified up front.
Calibration up front shortens everything downstream
The searches that finished near the median tended to share an opening. The hard conversations happened first: what the role must accomplish, what disqualifies a candidate, what a yes actually requires. That work felt slow at the time. It felt like delay before the real search began.
It was the opposite. Every hour spent on calibration up front bought back several hours downstream, because the team was no longer relitigating the criteria with each new candidate in the room. They had decided already. The decisions that stall slow searches were made before any candidate arrived to make them feel urgent and personal. Front-load the judgment and the rest of the process moves like it has somewhere to be.
The offer stage is where unmanaged processes lose weeks
The single most expensive failure in the dataset sat at the very end. A search can run clean for weeks and then bleed days at the offer, when an unmanaged process suddenly has to invent its terms in real time. Compensation gets debated internally while the candidate waits. A competing offer appears that closer attention would have anticipated. The candidate's real motivations surface for the first time at the worst possible moment to learn them.
The offer is a stage that has to be managed like every other, and ideally one you have been quietly preparing for since kickoff, because you already knew, from the calibration, what this person wanted and what it would take. Treat it as a formality you reach after the real work is done and it will cost you. The searches that lost weeks at the offer almost always treated it as an afterthought.
Some roles run longer, and that is correct
Not every role should be fast, and the median hides this. Clinical roles and dual-fluency roles (the people fluent in both AI systems and the regulated domain they deploy into) tended to run longer than pure software roles. That is not a process failure. It is the search behaving correctly in the face of genuine scarcity.
The mistake is forcing those searches to the median by lowering the bar until the pipeline fills. The fix is to expect the longer runway, specify the role even more carefully, and manage the decisions even more tightly, because on a scarce search there is less slack to absorb a stalled debrief or a fumbled offer.
Forty-three days is an artifact. We do not chase it. It is what tends to happen when the role is specified, the decisions are owned, and the offer is handled with the same care as the first conversation. Build the structure and the speed arrives on its own.