The same scene plays out in every recruitment team we meet. A role opens. The recruiter posts it and sources on LinkedIn and the job boards. Three weeks later, a spontaneous application comes in from a profile already seen six months earlier for another assignment, already evaluated, already rated positively at the time. Nobody remembered.
The talent pool is right there, it's full, and it stays invisible.
This is what we call the dormant talent pool: the share of your candidate database that has already been through an interview, already been rated, already been decided on, and that nobody will reopen for the next need because the information is unusable.
When you ask heads of recruitment what share of their talent pool actually gets used on new roles, the answer comes in two stages. A silence, then a range.
The talent pool isn't used 20% of the time. Recruiters reinvent the wheel with every new role.
We hear this estimate in a similar form in many of the teams we meet. It remains a field impression. No public study we could find puts a number on how much of the pool gets reused from one role to the next, and the tools don't measure it either. Yet everyone senses the same thing: most already-evaluated candidates never come back into the loop.
Public figures mostly tell you where hiring starts. In its study Executive recruitment practices 2026 (in French), Apec reports that 84% of French companies posted a job offer for at least one executive hire in 2025, and that 54% consider the posting the channel that most often found the candidate they ended up hiring. The default reflex is still the job ad, well ahead of the existing database.
Three technical reasons explain this waste.
The valuable information from an interview ends up in three places that talk to each other poorly: the free-text comment field of the candidate profile in the ATS, the recruiter's shared document (or notebook), and the spoken word, during the debrief with the manager. Six months later, you're looking for a cybersecurity profile available early in the year, OK with two days on site, who said they enjoyed ransomware topics. These three criteria exist somewhere in the database, but no tool has the material to find them again.
Many ATSs provide tags or structured skills. In theory. In practice, in a team running through thirty interviews a week, those fields aren't filled in correctly, or they're filled in with inconsistent vocabulary from one recruiter to the next. The filters exist, but they bring back nothing useful: "senior" according to Sophie doesn't mean "senior" according to Karim, and a search for "senior full-stack developer" returns two hundred results, of which a hundred and fifty are irrelevant and fifty are mislabeled. The five data entry rules that make a profile findable fix much of the problem at the source.
A recruiter isn't looking for a word, they're looking for a fit. "A profile who could replace Marc, more junior but with the same product instinct": that isn't a SQL query. As long as the candidate database can only be queried through full-text search or rigid filters, it stays closed to most of the questions a team asks day to day.
The subject sounds abstract. It becomes concrete the moment you put it in euros.
Take a team of six recruiters in a recruitment agency that opens twenty assignments a month. Let's assume that, on each assignment, two to three candidates already in the talent pool could have been called back instead of being sourced from scratch. What you lose every time:
- The sourcing time: under this assumption, two to four hours to rebuild an equivalent shortlist on LinkedIn and the job boards.
- The cost in candidate attention: a profile called back on the basis of a previous interview responds better, faster, and goes into the process more calmly than a profile approached cold.
- The risk of loss: while you're searching outside, the candidate already seen six months earlier is going to respond to another agency that did call them.
Over a year, for a modest team, we're talking about a hundred and twenty to two hundred and forty assignments where the talent pool could have taken the lead. At three hours of sourcing saved per assignment, the bill in recruiter time is far from trivial.
ATSs aren't to blame, they weren't designed for this. An ATS is first and foremost a pipeline tool: tracking applications by role, by status, by process stage. Its reason for being is "where does this candidate stand on this role," not "who in my database matches a need I haven't even formulated yet."
Talent pool search features often came as an afterthought, as a layer on top, with the keyword-search limits we described above. And when an ATS adds AI, it's usually to do role-to-CV matching: you have a role, you find the best CVs for it. Fine, but that doesn't solve the problem of unstructured interview data. At the matching stage, what's usable is what's in the CV. Not what was said during the forty-five-minute conversation.
That's the missing layer: making what you learned during the interview usable, instead of reinventing the phase that comes before it.
When interview data is structured properly and searchable in natural language, a recruitment team's day-to-day shifts on three fronts.
Instead of starting with LinkedIn, you start by looking at what you already have in the database for the need at hand. The talent pool doesn't replace external sourcing. It reduces its volume, since every profile found in the database is one less cold search.
A consultant met a year ago, one the team was keen on but whose timing didn't allow a placement, resurfaces on its own. Not by chance, their criteria match what you're looking for today. It's often the first effect the team notices.
A candidate called back on the basis of a previous interview gets a call that starts with "we met in March, you'd told us you were interested in topics X and Y, I have an assignment that fits." That tone, in a tight market, is nothing like a cold prospecting email.
Before talking about tooling, the diagnosis happens internally. Four questions to figure out where you stand.
- How many interviews has your team run over the past twelve months?
- How many candidate profiles have a structured interview comment, beyond a free-text note?
- How many searches against the internal talent pool are run per month? Many ATSs let you export it.
- What percentage of placed candidates came out of the internal talent pool rather than cold sourcing?
To go beyond these four questions, an ATS data audit establishes what the database can measure and what is missing.
If you don't have the answer to three of these four questions, that's already the main indicator: the talent pool isn't being managed.
You don't go from a dormant talent pool to an active one by switching ATSs. You do it by adding the layer that was missing: enriching the candidate profile at every interview in a structured format, then making the database searchable in natural language.
That's what Hirify works on. A Hub that plugs into your existing ATS, with no migration, and turns the interview hours you already run into data you can reuse six months later.
The operational method is detailed in a separate guide, to keep this one from getting heavy: see Reactivating a candidate pool in 5 steps.
It's the share of your candidate database that has already been interviewed and evaluated, yet nobody reopens for the next need. The candidates are in the ATS, but what the team learned about them sits in free-text comments, badly filled labels or recruiters' memories, so no search brings them back up.
No public statistic we could find measures this rate, and ATSs don't calculate it. In the teams we meet, most already-evaluated candidates are never called back for a later role. Public figures mostly show where hiring starts. According to Apec, 54% of French companies say the job posting was the channel that most often found the executive they ended up hiring in 2025. The only reliable figure for your team is the one you measure yourself, by comparing hires from the pool with total hires over twelve months.
The CNIL, France's data protection authority, recommends a reasonable period depending on the type of role, which should in principle not exceed two years from the last contact with the person, provided the candidate has been informed. It considers five years of retention to feed a talent pool excessive in principle. A dormant pool is therefore also a compliance issue, since files nobody reopens end up exceeding that limit.
Yes. The CNIL recruitment guide takes the example of an agency that builds a talent pool for its placement business with clients, and says it can be qualified as the data controller for that file. The same rules apply, with a legitimate purpose, information given to candidates and a limited retention period.
- In the teams we meet, most already-evaluated candidates are never called back for a later role. No public statistic measures this rate, which is why it's worth measuring in-house.
- The data exists, what's missing is structure: free-text comments, inconsistent labels, ill-suited search.
- The cost shows up as repeated sourcing time and already-qualified candidates lost to competitors.
- The answer is a layer added to your existing ATS, which structures the interview and makes the database searchable in natural language.