A role opens, and the recruiter starts a fresh search on LinkedIn, even though your team interviewed people for something close to it a few months ago, and every one of them is in your candidate database. The recruiter skips the database because a search there returns resumes and stage names, and none of the records say what the team learned about anyone: who impressed the panel, or why the finalist turned the offer down.
I think most teams try to fix that with better search, and I’d fix the records first. A search can only find what somebody wrote down. The candidate database becomes the first place recruiters look once each record holds what the team learned about the person, written the same way by everyone, at the moment they learned it.
I call that part of each record the signal layer: what the interviews showed and why each process ended, written down on top of the resume. Recruiters will mine a database like that first, which means searching it before anyone opens an outside channel, because that’s where the people your team already knows are kept.
What each candidate record has to hold.
Start with the record, because every later step depends on what one record can tell a recruiter who has never met the candidate. Most databases hold the first row of this table and little of the rest:
| What the record holds | What a later search can do with it |
|---|---|
| The resume and the contact details | Find people by past titles and employers, which an outside search can do as well |
| Each role they were considered for, and the stage they reached | Find people who got close on a similar role, ahead of people who left at the first screen |
| The evidence from each interview, filed under the competency it speaks to | Find people by what they showed the panel, such as a system they described building |
| Why the process ended, and which side ended it | Tell a candidate who chose another offer from one your panel turned down, before anyone writes to them |
| The next role they told you they wanted | Approach people about the role they asked for, instead of whichever role happens to be open |
| When your team last spoke to them, and whether they agreed to hear from you | Write to people who expect it, and leave out anyone whose permission has lapsed |
Every row below the first comes from your own team’s contact with the person, so no outside search can give you those rows. If you add one field this quarter, record why each process ended. A candidate who turned down your offer and a candidate your panel turned down need very different first messages, and the close reason is the quickest place for a recruiter to tell them apart.
Give every record the same shape.
One interviewer writes “strong communicator” in a note, and another rates communication four out of five. A search for communication finds one of them, because the two records were written different ways.
Agree on these fields once, and name one person to keep them:
- Competency names that match the ones on your scorecards, so interview evidence and scorecard ratings land under the same heading.
- A short, fixed list of close reasons, such as “chose another offer,” “withdrew,” “not ready yet,” and “turned down after the final round.”
- Stage names written the same way for every role family, so “reached final round” means one thing across the team.
- A free-text field under each competency, because the evidence itself rarely fits a dropdown.
Keep the required fields few. Require only what searches use, and I’d start with the close reason and the stage reached. Leave the rest optional until someone searches for a field and finds it empty, which is the moment you know the team needs it.
Fill records in while the work happens.
Most databases go stale because filling them is a separate job, booked for a cleanup day that keeps slipping. The fix is to tie fields to existing moments, ones your team goes through anyway, so the person who learned something writes it down while they still remember it:
- When an interviewer submits a scorecard, file their evidence on the candidate’s record under the competency it speaks to.
- When a process closes, have the recruiter choose the close reason before the stage can change.
- When a candidate declines an offer, note what they said they wanted instead.
- When someone asks not to be contacted again, mark it on their record that day.
Where interviews are recorded, file the evidence from the recording instead of from someone’s memory of the call.
Keep records current by using them.
A record goes stale in two ways. The person moves on, and the team forgets to write down what it learned the next time it spoke to them. A cleanup day fixes the first for a week and does little for the second, so update records on every contact.
Show the date of the last conversation on every record, so a recruiter knows how old the evidence is before trusting it. Interview evidence from two years ago still says what the candidate could do then, and the first call can confirm what has changed since.
When a recruiter writes to someone again, put the reply on the record, whatever it says. “Not now, ask me next spring” is one of the most useful lines a record can hold, and it’s lost if it stays in one recruiter’s inbox.
Merge duplicate records when a search turns them up. Two records for one person split the evidence between them, so the person looks thinner in both.
Search the candidate database before sourcing outside it.
Mining the database first means running a search of it before anyone opens an outside channel. Unless that search has a set place in the process, the recruiter goes back to the outside search they already know.
So search during the kickoff, before the intake call ends, using the profile the hiring manager gave you a minute ago:
- People who reached a late stage on a similar role, with the reason their process ended.
- People with interview evidence on the competency this role needs most.
- People who said they wanted a role like this one the last time they spoke to your team.
- Former finalists whose close reason was timing, such as “not ready yet.”
Go through what comes back while the hiring manager is still on the call, so the first conversation about the role is about people your team has already met.
Then keep a note of which candidates each search turned up and which of them moved forward. After a few roles, that note tells you which fields are earning their place, and which one was missing when a search came back empty.
Writing to the people a search finds is a separate job with its own risks, and this guide to candidate rediscovery walks through it.
Where Metaview fits in a candidate database.
The method drags in two places: getting what a candidate said onto the record, and searching for people by it later. I’d start with the search, because that’s where a recruiter decides whether the database was worth opening.
When a role opens, AI Sourcing takes a plain-language description of the person you want and looks for them on the web and in your connected ATS, plus your team’s earlier Metaview conversations. In the ATS it covers the candidates backfilled from the connection. Each candidate comes back with reasons attached. The agent doesn’t reach out to anyone or judge them, so a recruiter reads those reasons before anyone gets a message.
Reports lets you filter your team’s recorded interview data by fields such as interview stage and hiring manager, and pull up conversations by what came up in them. That covers the parts of the kickoff search that rest on what candidates said, while the close reason and the stage reached stay in your ATS. Reports also compares question coverage across interviewers and roles, so you can see which interviews asked about a competency at all:
- 1The filter at the top: backend software engineering interviews captured in the last month.
- 2One column per competency, here “Problem Solving,” “Coding,” “Collaboration,” and “Culture.”
- 3The “+” at the end of the header row, beside the competency columns.
For the fields your team never typed in, Reports can pull a data point you describe out of your interview data, such as the salary range each candidate gave. Metaview reads every transcript in your account to fill it in, back through your history and then for each new conversation.
- 1Each candidate beside the type of call it came from, from a recruiter screen to an intro chat.
- 2A new column, “Compensation Expectations,” added beside the candidates.
- 3Two rows where the column is still being filled in.
In a four-minute 2024 episode of the “Interview Data Masterclass” series, Siadhal Magos, who runs Metaview, does this kind of search. In it, he filters three months of conversations with sales candidates down to the ones who talked about closing large deals, then adds a column listing the companies each of them worked for before.
I’d copy his second step for any role you hire for often, because the companies your past candidates came from make a sourcing list your team already holds.
None of those searches can find what never reached the record. The Metaview Notetaker takes care of that part. It joins the call as a visible participant, and if each person on it has agreed to a recording, it makes a recording and a transcript, then arranges the notes in your team’s own note format. Your own process gathers that agreement.
Since Metaview captures every spoken word on the call, the evidence on a candidate starts from what they said, and a recruiter searching a year later for a skill they described can find them by their answers.
We built the #1 AI Notetaker for recruiting, but that was just the wedge. It gave us access to the most valuable data in hiring: your conversations”
Signs the database is still being skipped.
When recruiters keep skipping the database, look for the reason in the record before you look at the recruiters:
| What you see | Where the record falls short | What to fix first |
|---|---|---|
| Recruiters search outside before they search the database | The kickoff has no slot for the database search | Add the search to the intake call, and go through the results with the hiring manager there |
| A search returns names, but the records don’t say why each process ended | The close reason is optional | Require the close reason when the stage changes |
| One competency shows up under several names | The fields were never agreed on | Map each name to a scorecard competency, then merge them |
| A candidate replies that they asked not to be contacted | Permission lives outside the record | Mark the request on the record the day it’s made |
Go back to the note you keep from kickoff searches, because it’s the measure I’d watch. Once most entries name someone your team has already met, with a reason to call them on the record, the database has become the signal layer recruiters mine first.
Make your candidate database the first search on every role.
See Metaview AI Sourcing search your ATS and every past Metaview conversation from one prompt, and explain why each candidate came back.
Frequently asked.
How is a candidate database different from an ATS?
An ATS moves candidates through the roles that are open now. A candidate database is the record of everyone your team has met, and it often lives inside the ATS. The difference is when it gets read: a database search usually runs long after the process that created each record has closed.
Should sourced prospects who never applied go in your candidate database?
Yes, with a tag saying where each record came from. A sourced profile holds public information until someone on your team speaks to the person, so the tag lets a recruiter see at a glance whether the team has ever met them.
Should you keep candidates rejected at the application stage?
Keep the record and its close reason, and expect little from it. An application turned down before any conversation holds a resume and no interview evidence, so read those records after the ones for candidates your team has spoken to.
How long can a candidate database keep someone’s details?
Your privacy lead sets the period from the law in each place you hire and from your company’s own policy. Tell candidates what you keep when they enter a process, and delete each record when the period ends.
Who should own the candidate database?
Recruiting operations, where you have the role. One owner agrees on the fields and the close reasons and checks them against the searches recruiters run. Recruiters and interviewers fill the records in, because they’re the people who learn what goes in them.