The talent pool nobody searches before paying to source

Because searching your own database has become slower and less reliable than searching outside it, and because nobody is responsible for doing it before money is spent. The paradox is that the talent pool is now the leading source of sourced hires: according to Gem, 44.0% of them in 2024 involved a candidate already in the company’s CRM or ATS, up from 29.1% in 2021. Teams are buying back, one subscription or job ad at a time, profiles they already own.

A better search tool will not fix this on its own, although one is needed. It is fixed by deciding that the internal search comes before any external spend, and then by knowing which profiles in the pool deserve a call and which deserve to be left alone. The second half is the one most often got wrong, and it goes wrong in the other direction: through the reactivation campaign sent to everyone.

How many of your hires already come from your own database?

Probably more than you think, and you can measure it yourself in an hour.

Gem’s figure has limits that need stating before it is used. It is US data, drawn from the customers of a vendor that happens to sell a candidate rediscovery tool, and the article gives neither the sample nor the method. It shows a direction rather than a value, and that direction matches what most agencies say once they have run the numbers: a large share of what they place was already somewhere in their files.

The calculation is simple. Take your placements from the last twelve months and, for each one, the date the person entered your database. Anyone who was there more than three months before the role opened is a placement from the pool, even if you found them by chance or because the candidate got back in touch on their own. The second number is less comfortable: of those placements, how many were preceded by paid sourcing on the same role? That is what you spent to find what you already had.

Why does nobody search the pool before sourcing?

Because external search returns fresher results, faster, and internal search disappoints often enough that people stop trying it.

The first reason is mechanical. A keyword search in an ATS finds what was written the way you write it, and a developer entered three years ago as a “software engineer, applications” does not come up when you type “back-end developer”. The profile exists, it fits, and it is invisible. Cobalt has set out what a CV database nobody uses actually costs, and why keyword search is the first cause, so we will not go over it again here.

The second reason is freshness. A contact database loses about 30% of its freshness every year, and that figure, which we went into when looking at the data you hand to an agent, hits exactly the fields that decide whether a call is worth making: current employer, availability, contact details. A recruiter who opens three out-of-date records in a row concludes the database is worthless, when what it mostly needs is to be checked at the moment it is used. The profile on a professional network, meanwhile, was updated by its owner last week, and that asymmetry explains most of the reflex.

The third reason is organisational, and it is the most stubborn. Nobody owns the pool. External sourcing has a budget, a tool and someone whose job it is; internal search has none of those, and the memory of who was right for what leaves with the recruiter who leaves the firm. What stays in the ATS is the record, without the conversation that gave it meaning.

Which candidate in the pool actually deserves a call?

One with whom a real conversation took place, who was not turned down on fit, and about whom something has changed since. All three conditions matter, and the third is the one people forget.

Profile in the pool	What it is worth
Finalist turned down for reasons outside their control	The best: budget frozen, role filled by a narrow margin, client changed its mind. The assessment is done.
Candidate who declined because of timing	Very good if the delay they gave has passed. They told you when to come back; you only have to remember.
Profile with a recent sign of change	Good: end of an assignment, a reorganisation at their employer, tenure reaching a milestone.
Candidate turned down on fit	Weak, unless the role is genuinely different. Calling them about a similar role looks like an oversight.
Imported profile or application with no exchange	The weakest: you know no more than about a stranger, with older data.

The table says something that is hard to see until it is written down: the value of a profile in the pool lies almost entirely in the conversation that took place, and hardly at all in the stored CV. A finalist turned down eighteen months ago has been assessed, sometimes referenced, and knows how you work. A record imported in bulk is just an address, and often a wrong one.

The ranking is done with a sentence, not a score. “Finalist on this client’s June role, turned down because the budget was frozen, current assignment ended in September according to their profile” can be checked and challenged; a match score of 82 says none of that, which is why we hold that the score is the wrong instrument for this work as for the rest.

Why does mass reactivation burn the pool?

Because it is paid for twice, first in sender reputation and then with the candidates, and both costs outlast the campaign.

The first cost is technical and little known among recruitment teams. A campaign to addresses three years old produces bounces and spam reports, and Google requires senders to keep their complaint rate below 0.3%, recommending they stay under 0.10%. Past that point the campaign is not what suffers; the whole domain is, and your recruiters’ individual emails, the ones that matter, start landing in junk folders. A single badly judged reactivation can damage weeks of normal correspondence.

The second cost is human, and it is worse. A good part of your pool is made up of people you never answered: 53% of candidates were left without a reply by an employer in 2026, and that silence comes from an overloaded queue rather than carelessness, but the candidate cannot tell the difference. If the first message they get from you since then has clearly gone to a thousand people, it confirms what they already thought. You had a disappointed candidate; now you have one who will never reply again.

This is the candidate-side counterpart of what we wrote about the clients nobody calls: generic reactivation is the shortest route from dormant to lost. You only get to reactivate a pool once with a message like that, and that once is wasted.

How do you reactivate a talent pool without losing it?

In small batches, on a specific role, with a message that says why this person and why now, and that acknowledges what happened last time.

The rule that shapes the rest fits in one sentence: you search the pool before every paid sourcing effort, every time, and you only contact people from the pool for a real role. The aimless “keeping in touch” message is exactly the one people learn to ignore. An open role, on the other hand, gives a reason, and a reason gives the message the one thing that sets it apart from a campaign.

The message itself picks up where the relationship left off. If the person was a finalist, say so; if you never got back to them, say that too, in one line and without a paragraph of apology. That reminder costs little and changes how everything else is read, because it proves somebody opened the record before writing.

One constraint we mention without going into, because it deserves an article of its own: a pool does not last forever, the time for which a candidate’s data may be kept is limited, and a profile kept beyond it is no longer an asset but a compliance breach.

What can an agent do with the pool?

What nobody has time to do at the right moment: search before spending, check before proposing, and remember the conversation that took place.

Systematic search is the ideal case of a task to delegate first, because its output can be checked at a glance. A list of eight profiles with, for each one, why it is there and what could not be verified takes five minutes to review. The freshness check is done on those eight records and not on the eight thousand, when they are needed, which is the only clean-up strategy that holds up against a database that is ageing all the time.

The message stays on the human side of the line. The agent can draft it, drawing on what the record says about the last conversation; a person reads it, adjusts it and sends it, because it is an outbound message and it commits the whole relationship. That is how we have Balt work on a pool: it searches before any sourcing, tells you who it found and why, and nothing goes out without you.

Does the pool replace external sourcing?

No, and claiming otherwise would mean selling a saving that does not always exist.

A pool reflects the roles you have had, not the ones you will have. A skill that did not exist two years ago, a new client sector, a city where you have never hired: in all these cases internal search will return little, and that is to be expected. The pool also ages, and a database that is never replenished ends up holding nothing but people who have changed careers.

So the right framing does not set the two against each other. You search the pool first, because it is free, quick when search works, and the conversation has already happened there; you source outside next, for what the pool does not have. The order is the only change, and it is enough to remove the most absurd expense in the trade: paying to find someone you already knew.

The next question is about the people you have already placed, who are the quietest of all: the consultant on assignment nobody calls leaves without warning, and never comes back through the pool.