One number should make every recruitment agency leader stop and think. According to Gem's annual report, based on 140 million applications and 1.3 million hires, the share of sourced hires coming from candidates already in the database rose from 29% in 2021 to 44% in 2024.
In other words: nearly one sourced hire out of two requires no headhunting at all. The candidate is already there. They've already been met, qualified, sometimes even presented to another client.
The problem is that most agencies don't actually know what their own database contains. CVs pile up for years in a CV database that nobody queries. Talent pool software exists precisely to turn that dormant stock into a usable asset.
What to remember
- 44% of sourced hires come from candidates already in the database (up from 29% in 2021, source: Gem). Nearly one hire out of two therefore requires no external headhunting.
- CV database, talent pool and pipeline are three different things: the CV database stores the past, the talent pool prepares the future, the pipeline manages the present.
- Four causes make most talent pools useless: no segmentation, stale data, a relationship that's never maintained, ineffective internal search.
- The method comes down to four steps: structure profiles (skills, seniority, mobility, interview notes), segment finely, maintain data freshness, reactivate silver medalists first.
- GDPR is a tool-selection criterion: the CNIL sets 2 years after last contact as a reasonable maximum, interview notes included. The tool must allow exhaustive deletion in a single action.
CV database, talent pool, pipeline: three different things
These three words get used as synonyms. They are three distinct tools, with three purposes and three time horizons.
- The CV database stores. It's a database of received CVs, sorted and kept. It answers the question "what have we received?" Without processing, it's passive by nature.
- The talent pool anticipates. It's a set of qualified, segmented and maintained profiles, built for future needs. It answers the question "who could match an upcoming role?"
- The pipeline manages the present. These are candidates engaged in an active process, with a status and a next step. It answers the question "where do my open applications stand?"
The difference lies in usage and timing. The pipeline handles the moment, the talent pool prepares the future, the CV database preserves the past.
CV database software that only stores will therefore leave you halfway there. What gives a database its value isn't the number of profiles it holds, it's the ability to surface the right one at the right time. That's the difference between archiving and using.
Why do most talent pools serve no purpose?
The observation is almost universal. 68% of companies say they keep the CVs they receive, but fewer than 30% actually re-contact candidates. The database exists, it grows every month, and it produces nothing.
Four causes explain this waste.
1. No segmentation
This is the main cause. A talent pool of 5,000 profiles with no segmentation is a talent pool of 0 usable profiles. Faced with an urgent role, nobody has time to scroll through an undifferentiated list. Without a system of tags, custom fields or job categories, the database stays a pile of CVs rather than a working tool.
2. Data goes stale fast
A profile entered two years ago no longer reflects reality. The candidate has changed jobs, earned a certification, moved, or become available again without your knowledge. A frozen talent pool loses its value within 2 to 3 years, right at the moment its size would start to make it interesting.
3. The relationship is never maintained
A candidate met once and never followed up with isn't in your talent pool: they're in your archives. The difference is played out in the relationship. A profile with whom contact has been maintained will reply; a profile forgotten for three years will treat your message as a cold outreach, with the response rate that comes with it.
4. Internal search is ineffective
Even a well-filled talent pool is only usable if you can query it quickly. When search relies on exact keywords, half of the relevant profiles never surface. This is the topic we covered in detail in our article on sourcing software: the search method directly determines the value of the database.
These four causes compound each other, and the result is always the same: faced with a new role, the reflex is to go look outside. Not out of laziness, but because it's objectively faster than digging through your own database. That's the calculation a good tool has to flip.
The 4 steps to turn your database into an asset
Moving from a dormant CV database to a usable talent pool doesn't mean starting over. It means treating the database as an asset, with a method, and talent pool software able to support it.
Step 1: structure
Before segmenting, every profile needs to carry the right information. The exploitable minimum for an agency:
- Actual skills, not just the job title
- Seniority and level of responsibility
- Geographic mobility and any constraints
- Salary expectations and the type of contract sought
- The history of exchanges: who met this candidate, when, and what came out of it
This last point is the most often neglected, and yet it's the one that sets an agency apart from a job board. Your consultants' interview notes are worth more than the CVs, because they contain what no document says.
Step 2: segment
A useful talent pool is a segmented talent pool. The segmentation axes that work for an agency:
- By job and specialty, at the level of granularity of your recurring roles
- By seniority level, because approaches differ radically
- By availability: employed, open to opportunities, actively searching
- By commercial value: the profiles matching your most profitable or rarest roles
The goal is simple: finding "every CFO with 10+ years of experience, willing to relocate to Lyon" should take less than 30 seconds. If it takes longer, the segmentation isn't good enough.
Step 3: maintain
A talent pool that doesn't move dies. The minimum rhythm to keep up:
- An annual update of strategic profiles
- A follow-up at 6 months after an interview, even with no role to offer
- Archiving unreachable profiles after 24 months, which aligns with GDPR obligations
- Verifying contact details before every outreach campaign
Maintenance isn't just a matter of data freshness. It's also what makes a candidate reply when you finally reach out.
Step 4: reactivate
This is the step that generates the return on investment. The profiles to target first:
- "Silver medalists", candidates who came second on a mission. They were judged good, they already know your agency, and they're often still open.
- Profiles who have moved into active search since your last exchange
- Candidates whose background matches a role you didn't have at the time
In summary:
| Step | Question to ask | Sign it's working |
|---|---|---|
| Structure | Does every profile carry skills, seniority, mobility and interview notes? | A candidate record reads without opening the CV |
| Segment | Can I isolate a precise population in a few seconds? | A targeted search takes less than 30 seconds |
| Maintain | Is my data less than 12 months old? | Contact details are reliable before every campaign |
| Reactivate | Am I re-contacting my silver medalists? | A share of your placements comes from the talent pool, not headhunting |
GDPR: what you're allowed to keep, and for how long
This is the point most talent pool articles leave out, even though it dictates how you must build your database. A candidate talent pool isn't an address book: it's personal data processing, with the obligations that come with it. (This section reflects French regulation — the CNIL is the French data protection authority — the applicable rules in your own jurisdiction may differ.)
The 2-year rule
The CNIL, France's data protection authority, considers a retention period of 2 years after the last contact with the candidate to be a reasonable maximum. Beyond that, retention must be justified by a real necessity, or rely on the person's explicit consent.
In practice, for an agency that wants to build a lasting talent pool:
- Inform the candidate at the point of collection: purpose, duration, rights they have
- Collect distinct consent if you want to keep the profile beyond the standard period. A dedicated checkbox, with clear wording such as "retained for 2 years for future opportunities"
- Keep proof of that consent (timestamp, text shown)
- Allow withdrawal as easily as it was given
What exactly the obligation covers
A point that's often misunderstood: interview summaries are personal data too. They follow the same rules as CVs and must be deleted at the same time if the candidate exercises their right to erasure. Your consultants' notes, email exchanges, evaluations: all of it falls within scope.
This means a deletion request exercised by a candidate must be handled without delay and exhaustively, which requires that their data isn't scattered across four different tools.
Why it's a tool-selection criterion
Compliance isn't just a legal constraint, it's a technical one. Good talent pool software should let you:
- Automatically track retention periods and flag profiles to archive
- Purge a profile entirely, notes and history included, in one action
- Track consents and their date
- Guarantee data location, a point that matters for a company operating under GDPR
This last point deserves attention: hosting a French candidate talent pool with a vendor outside the European Union complicates compliance. At Marvin, data is hosted in Europe and GDPR compliance is built in by design, not bolted on afterward.
Periodic cleanup, finally, isn't just an obligation: it's also good database-quality practice. A talent pool purged of dead profiles is a talent pool that's faster to query and more reliable to use.
Marvin, the talent pool software that maintains your database
The four steps above take time. And that's exactly what's missing at an agency. Our approach is to automate what can be automated and make the rest instant.
A talent pool that updates itself
That's the role of our Talent Refresh agent. It works continuously on your database: it enriches profiles, detects changes in status, and specifically flags candidates who've switched to "Open to Work."
In practice, you no longer discover a profile is available by calling them at random. You're notified when they become reachable again, at the moment the information has value. The "maintain" step stops being a quarterly chore and becomes a background process.
A search that queries everything, not just CVs
With Scout, semantic search covers your entire database: CVs, interview notes, exchange history, imported LinkedIn profiles. This is what finally gives value to your consultants' work.
A note jotted down after an interview three years ago, something like "very good at crisis management, will look to move in two years," becomes searchable. Your agency's knowledge stops being locked inside the memory of whoever ran the interview.
Segmentation without manual tagging
Because profiles are analyzed semantically, you don't need to have manually tagged every candidate to find them. You describe the population you're looking for in natural language, and it's built on demand. Segmentation becomes dynamic instead of frozen into categories decided in advance.
The talent pool connected to the rest of the business
The talent pool doesn't live in isolation: it shares the same data as Desk, our ATS and CRM. A reactivated candidate goes straight into a process, the relationship is tracked, and compliance is managed in the same place. A deletion request is handled in one action, across all of the candidate's data, notes included.
At a glance, here's what changes:
| Classic CV database | Talent pool run with Marvin | |
|---|---|---|
| Nature of the database | Passive CV stock | Continuously queryable asset |
| Updating | Manual, whenever someone thinks of it | Automatic via Talent Refresh |
| Search scope | CVs only | CVs, interview notes, history, LinkedIn |
| Segmentation | Manually entered tags, frozen | Dynamic, in natural language |
| Availability detection | At the mercy of random follow-ups | Flagged when the profile becomes open |
| GDPR compliance | Scattered across several tools | Centralized, deletion in one action |
The gain is twofold. Our clients replace an average of 5 to 7 tools with a single platform and free up around 7 hours per week per consultant. But above all, they finally make use of an asset they already owned: every profile met over the years becomes a presentable candidate again the day a role matches them.
Your database is worth more than you think
Let's go back to the number we started with. If 44% of sourced hires now come from candidates already in the database, that means an agency that doesn't use its talent pool is working twice as hard for the same result. It's hunting outside for what it already owns.
Moving from a CV database to a talent pool isn't about the size of the database, it's about four habits: structure, segment, maintain, reactivate. And about talent pool software that makes those habits sustainable day to day, because no agency has the time to manually maintain several thousand profiles.
It's also a matter of company value. A recruitment agency or an executive search firm selling its business is selling a client portfolio and a qualified candidate database. A CV database full of stale CVs is worth nothing; a segmented, up-to-date and compliant talent pool is a real asset.
Want to know what your database actually contains? Book a demo: we'll query your talent pool under real conditions to see how many profiles match your open roles, and show you how migrating your data from your current tool works.
Frequently asked questions about candidate talent pools
What's the difference between a CV database and a candidate talent pool?
The CV database stores, the talent pool anticipates. A CV database is a database of received CVs, sorted and kept: it answers the question "what have we received?" A talent pool is a set of qualified, segmented and maintained profiles built for future needs: it answers "who could match an upcoming role?" The difference lies in usage and timing, not content. A CV database becomes a talent pool once it's segmented, kept up to date and actually queried.
How long can you keep a candidate's CV?
The CNIL, France's data protection authority, considers a duration of 2 years after the last contact to be a reasonable maximum. Beyond that, you need the candidate's explicit consent, collected separately and provable. Note: this rule also covers interview summaries, which are personal data in the same way as CVs. A candidate exercising their right to deletion must have all of their data erased, notes included.
Is there a free CV database?
Free solutions exist, generally limited in number of profiles, search features or duration. The real cost, however, isn't there. A free CV database with no fine-grained segmentation, no automatic updating and no compliant data deletion will waste more time than it saves. For an agency whose business depends on its candidate database, the question isn't the price of the tool but the number of placements the database allows you to make.
How do you reactivate a dormant talent pool?
Start with the profiles most likely to respond: "silver medalists," candidates who came second on a mission, who already know your agency and were judged good. Then profiles whose situation has changed since your last exchange. Three principles for re-establishing contact: personalize the message by recalling the context of your last exchange, bring a concrete reason (a specific role, not a generic follow-up), and verify contact details before sending. A talent pool is reactivated in targeted waves, not through a mass campaign.
How many profiles does a talent pool need to be useful?
There's no magic threshold, and size is a poor indicator. A talent pool of 500 well-segmented, up-to-date profiles is more useful than a database of 10,000 stale CVs. What matters is the share of usable profiles: those with fresh data, documented skills and maintained contact. The right test is simple: faced with a typical role at your agency, how many relevant candidates does your database surface in under a minute?
