Time-to-fill: 18 days for one firm, 81 for another

Because the gap does not measure the market, it measures dead time. Median time-to-fill for an IT consultant placed with a client stands at 45 days in 2026, but the top decile places in 18 days and the bottom in 81, and these firms recruit the same profiles, in the same year, in the same country.

A ratio of 4.5 between the extremes of one population rules out nearly every comfortable explanation. Pressure on skills is shared, salary bands look alike, the job boards are the same. What differs is how a file is held between two useful actions, and that is good news, because it is the only variable on that list that belongs to you.

What exactly does the spread say?

It says the median is the least interesting figure in the series. A median of 45 days invites you to compare yourself to an average and feel reassured; a spread of 18 to 81 days invites you to ask what the fastest ones are doing, which is a far more useful question.

The gap resists structural explanations. You might assume the fastest are the largest, with deep benches and dedicated teams, but the study producing these figures shows instead that the most fragile segment is firms of fifteen to fifty consultants, with turnover at 23.4% against a median of 18%. Size plays a part, and not in the direction that would suit everyone.

You also have to set aside the idea that the fast ones are trading quality away. A placement made quickly and badly shows up within three months, in turnover and in early terminations, and that is not what the segments concerned show. Which leaves the process.

Where do the 45 days actually go?

Almost all of the delay is waiting, and almost none of it is work. This is the finding that most surprises teams when we decompose a real file with them, because each of them remembers working hard.

Add up the time spent on useful tasks in one placement: qualifying a requirement with the client, sourcing, calling four candidates, assembling a skills dossier, arranging an interview, giving feedback. You get hours, spread across a dozen distinct acts. The rest, meaning the vast majority of the 45 days, happens between those acts, in intervals where the file belongs to nobody: the client has not replied, the candidate is thinking, the follow-up was due Tuesday and nobody remembers.

This diagnosis mirrors the one we make on the candidate side, where the median response time reaches 6.7 days while interest breaks around the seventh. The consequence is the same on both sides: it is not slow work that costs you, it is the latency between two actions, and it is paid in candidates who went elsewhere.

What do the top-decile firms do?

The study does not say, and I am not going to credit the fastest with methods they may not have. What I can describe is what we observe in teams that hold short delays, and it is remarkably unspectacular.

They have no magic bench. What they share are three habits that cost nothing and that nobody mentions on a conference stage. The first is that an open file always carries a dated next action, including when that action is to wait: “chase the client on the 12th” is an action, “with client” is not. The second is that the follow-up goes out on the planned day even when it has nothing new to say, which sounds impolite and turns out to be the opposite. The third is that the negative answer goes out fast, because a closed file frees a slot in the pipeline and in somebody’s head.

What the three have in common is that none of them asks anyone to work more. They ask you not to depend on one person’s memory on a busy Tuesday, which is exactly the kind of constraint a team holds for six weeks and then lets slip without noticing. Which is also why discipline alone does not explain the gap durably: the slow teams know these three rules, they were applying them two years ago.

Why dead time appears on no dashboard

Because it is charged to nobody, and that is a structural property rather than a tooling shortcoming.

Working time is declared by whoever spends it: it shows up in a report, in an activity log, in a conversation. Dead time has no owner, therefore no line, therefore no alert. A file that has not moved in eleven days produces no event, whereas a file being worked on produces one at every step, which gives you dashboards where the only invisible thing is precisely the main one.

It is the same asymmetry that distorts savings calculations on assisted drafting. Time saved is measured by whoever produces and gets declared readily, while time lost disperses among those who rework it and appears nowhere, which is enough to turn a transfer of load into a headline saving. Time-to-fill suffers the temporal version of the same bias.

What compresses, and what does not

The distinction deserves stating bluntly, because this is where most of the sector’s promises turn dishonest.

What does not compress: the client’s decision, which depends on their own committee and their own trade-offs; a candidate’s deliberation, who is entitled to take a week; a notice period. No technology acts on those three, and a vendor implying otherwise is selling a brochure.

What does compress is everything depending on somebody’s memory. The follow-up due Tuesday that goes out on Tuesday, the file picked back up at the appointed hour rather than the day you stumble on it, the client update sent on Friday without them having to ask. It is unglamorous to describe and it is where the gap between 18 and 81 days sits, because those pick-ups repeat ten times per file and across the whole active load.

It is also why we advise starting a delegation with what is verifiable rather than with what takes the most time. A follow-up that went out on time can be checked at a glance, and it is exactly the kind of task whose cumulative effect shows up in an overall delay.

What to measure instead of time-to-fill

Time-to-fill is a good outcome indicator and a poor steering one, for a simple reason: it does not decompose. When it worsens by ten days, it tells you neither where, nor since when, nor because of what.

The indicator I would recommend to a services firm’s management is the delay between two actions on the same file, taking its maximum rather than its average. It points at a specific place, it is corrected the same week, and it has the rare property of degrading before the outcome rather than after. A file whose longest interval goes from three to eleven days warns you a month before the median moves.

The second is the share of files with no dated next action. It is a hygiene indicator rather than a performance one, it reads in ten seconds, and it correlates strongly with what teams call among themselves the files that drift. A dated action on every open file does not guarantee the placement, it only guarantees that nobody is waiting in the belief that someone else has it, which is already most of the difference.

Which leaves the cost of all that time, calculated differently when the consultant is already on your payroll waiting for an assignment rather than for an employment contract. That is the subject of the bench as a staffing problem, where the same lost day carries a direct price.