Refolk
FrameworkRecruiting and sourcing

The Candidate Identity-Drift Read: Proceed, Step Up, Pause, or Withdraw

Score any mid-funnel identity flag on five dimensions and reach the verdict a colleague would, with the next step, its deadline stage, and neutral candidate wording.

23 min readLast reviewed November 13, 2026Read as Markdown

Key takeaways

  • Greenhouse Real Talent flags about 1 in 4 candidates for review but fails only about 1 in 33, so a single identity flag is roughly eight times more likely to clear than to fail.
  • Score identity drift on stable facts such as project details, timeline, contact channel, and shipping address, never on delivery, accent, or name; a new headset or an AI-polished answer is noise.
  • An anchor found before outreach and re-tied through a channel the candidate did not supply is the only identity evidence the candidate could not shape, because background checks can be pre-tested.
  • Fixed verdict rules: withdraw on a contradicted anchor or two dimensions at 2, pause on exactly one dimension at 2, step up on a drift sum of 2 or more, otherwise proceed.
  • In Refolk's index there are 7.8 US technical recruiters or sourcers for every US background-screening specialist, so on most commercial teams the recruiter carries the identity check.
  • In Checkr's manager survey, 28% of managers knowingly advanced a candidate they suspected of dishonesty, which is why the verdict rule has to be fixed before any flag appears.

This is the scoring model for one repeated call: a candidate you sourced or screened now looks or sounds different between rounds, writes from a new email or number, or asks to ship the laptop somewhere new, and you have to decide today whether to proceed, step up, pause, or withdraw. It is written for in-house recruiters, sourcers, talent leaders, and founders doing their own hiring. You score the flag on five dimensions, apply fixed verdict rules, and leave with a next step, the stage it must happen at, and wording that does not accuse an honest person.

What counts as an identity-drift flag?

An identity-drift flag is an observable change, between first reply and day one, in who you are talking to, how you reach them, or where their equipment and pay are going. A change in how an honest person presents (sound, lighting, polish) is noise, and this model is built to keep the two apart.

The signals that federal advisories tie to proxy and fake-identity hiring are specific. The FBI's public service announcements (PSAs) describe:

  • Background noise that sounds as if the applicant is surrounded by others doing similar work.
  • Face-swapping during video interviews.
  • VoIP numbers and email addresses reused across multiple resumes.
  • Changes in address or payment platform during onboarding.
  • Address changes after hiring but before laptop delivery, per a secondary summary of an FBI advisory.

The State Department adds the one that matters most for a mid-funnel read: the individual a hiring official interacts with "may change depending on the time of day," because these teams rotate. That is why comparing one round with the next catches what a one-off ID check misses.

Ordinary noise looks like this:

  • A new headset, camera, or lighting setup between rounds.
  • Coached or rehearsed answers.
  • AI-polished writing. Gartner reports that 4 in 10 candidates use AI on their applications.
  • VPN use, which commercial verification tools flag and which plenty of honest candidates trigger.
  • A genuine move to a new address.

The test that separates them: drift changes a stable fact (who answers, which channel reaches them, what they know about their own work history, where the equipment goes). Noise changes delivery. Accent, name, and perceived origin are neither, and they never enter the score.

How common is candidate identity fraud, and why is one flag weak evidence?

Identity fraud is rare per candidate but common per interviewer, so any single flag on any single candidate is far more likely to be noise than fraud. A verdict needs corroboration across more than one dimension.

The published measures count different things and are not interchangeable:

SourceWho was askedMeasureFigure
Gartner candidate survey (n=3,000)CandidatesAdmitted interview fraud6%
Checkr manager survey, earlier wave (n=3,000)US managersEver interviewed a fake identity31%
Checkr manager survey, later wave (n=3,000)US managersEver interviewed a fake identity40%
Greenhouse Real Talent (10.2M+ checks)Candidates checkedFlagged / failed25% / ~3%

The 6% and the 40% do not conflict. The first is a per-candidate rate and self-reported, so treat it as a floor. The second measures a manager's exposure across every interview they ran, and it rose 9 points between survey waves. Most interviewers will eventually meet a fake, while the individual candidate in front of you is still probably genuine. The same Checkr work found 67% of managers had encountered an interview proxy: 34% confirmed one and another 33% strongly suspected one. Half of that headline figure is suspicion, not confirmation.

Flags outnumber failures about 8 to 1

  1. Candidates checked
    10.2M+

    All Real Talent checks run

  2. Flagged for review
    1 in 4

    Includes data mismatches and VPN use

  3. Failed verification
    1 in 33

    A failure does not auto-disqualify

In Greenhouse Real Talent's checks, one candidate in four is flagged for review but only about one in 33 fails verification outright.

The flagged-to-failed ratio is about 8.25 to 1. Greenhouse's "Flagged" status explicitly includes information mismatches and VPN usage, and Greenhouse states that even a failed verification does not automatically disqualify a candidate. If the tool built to catch this does not treat its own flag as a verdict, a recruiter should not treat one raised eyebrow as one.

The five dimensions and how to score each

Score every flag on the same five dimensions, each from 0 to 2: anchor strength, channel continuity, cross-round consistency, logistics drift, and role access risk. The first four measure drift; the fifth sets how much drift the role can tolerate and which step-up check applies.

The anchor is the centre of the model. An anchor is a public record of the person that you found independently before outreach: a professional profile, a commit history on the public GitHub graph, a prior-employer contact. A re-tie is the act of linking the person now in your process back to that anchor through a channel the candidate did not supply.

Dimension0: clean1: ambiguous2: drift
Anchor strengthPre-outreach anchor, re-tiedNot yet re-tied, or thinIndependent channel contradicts
Channel continuitySame email and numberNew channel, plausible reasonDisowned, or reused elsewhere
Cross-round consistencyStable facts matchOnly delivery differsStable facts contradict
Logistics driftMatches ID addressMove, documents offeredChange, documents refused
Role access riskTier 0: no sensitive accessTier 1: code, internal systemsTier 2: production, admin, customer data

One scoring rule is what makes two recruiters land in the same place: a 2 requires a written, observable fact that contradicts another written fact. "He seemed different" is a 1 at most. "In round two the candidate did not recognise the payment migration he described in detail in round one" is a 2.

Anchor strength

  • What it proves: the identity in your process existed, did real work, and is reachable somewhere the candidate does not control. A sourced candidate starts strong, because the sourcer found the anchor before the candidate knew you existed.
  • How it lies: the anchor itself can be rented. The FBI says facilitators provide access to professional networking accounts. Check that the anchor's history was active before the role opened, and that it is reachable outside the contact details the candidate gave you.
  • Score: 0 if found before outreach and re-tied; 1 if thin, recent, or not yet re-tied (most inbound applicants start here); 2 if an independent channel contradicts it.

If you sourced the candidate in Refolk, the profile and public GitHub history came back before you wrote a word to them. Save that link on the candidate record the day you send outreach, because it is the one piece of evidence nobody in the process shaped.

Channel continuity

  • What it proves: the person replying now is reachable on the same line as the person who replied first.
  • How it lies: someone who changed phones has an honest new number. In the other direction, a "prior employer" reached on a number the candidate supplied loops straight back to the fraud.
  • Score: 0 for unchanged channels, or a change confirmed through the anchor; 1 for a new email or number with a plausible, unconfirmed reason; 2 if the anchor channel disowns the new one, or the number or email appears on other resumes.

Cross-round consistency

  • What it proves: the same person, with the same memory of their own work, is in each round.
  • How it lies: delivery varies for honest reasons. A new headset, different lighting, or an AI-polished answer scored as drift is the most common false positive.
  • Score: compare project specifics, timeline, team size, systems named, and reasons for leaving. 0 if consistent; 1 if only delivery differs; 2 if a stable fact contradicts, if you hear the room-of-others noise the FBI describes, or if a consented comparison image does not match.

Logistics drift

  • What it proves: equipment, pay, and access are going to the identity you verified.
  • How it lies: honest people move, often between offer and start.
  • Score: 0 if shipping and payment match the ID address; 1 for a change with documentation offered or pending; 2 for a new address or payment platform with documentation refused, or a request to ship to a third party. The FBI's own remedy for an address change is more documentation, not withdrawal.

Role access risk

  • What it proves: nothing about the candidate. It measures what a wrong verdict costs.
  • Score: tier 0 for roles with no production, customer data, or admin access; tier 1 for code repositories and internal systems; tier 2 for production, admin, finance, or customer data, and for contract IT work, which the FBI calls a common route for North Korean IT workers into employment. Set the tier when the requisition opens, never after a flag.
Drift flag log entry
Candidate ID:
Requisition and role tier (0, 1, 2), set at req open:
Stage when flag raised:
Logged by / date:
Observable signal (what was seen or heard, verbatim where possible):
Stable fact it contradicts, if any (quote the earlier note):
Pre-outreach anchor (link, date found, who found it):
Independent re-tie channel (how the number or address was found):
Scores: anchor _ / channel _ / consistency _ / logistics _ / role tier _
Drift sum (first four): _    Count of 2s: _
Verdict:
Next step, owner, and the stage it must happen before:
Same check applied to all candidates in this tier and stage? (yes / no, log ref):

Complete it the same day; record what was seen or heard, never what you inferred about the person.

How the score becomes a verdict

Add the first four dimensions into a drift sum from 0 to 8, count how many sit at 2, and apply the first rule that matches. Role tier changes the verdict only at the line between Proceed and Step Up, and it decides which step-up check you run.

The rules, in order:

  1. Withdraw if anchor strength is 2, or if two or more drift dimensions are at 2.
  2. Pause if exactly one of channel, consistency, or logistics is at 2.
  3. Step Up if no dimension is at 2 and the drift sum is 2 or more, or if the drift sum is 1 on a tier 2 role.
  4. Proceed otherwise: a drift sum of 0, or 1 on a tier 0 or tier 1 role.

A Pause is temporary. Run the re-tie, request the documentation, then rescore. If the 2 drops to a 1 or a 0, the new score sets the verdict. If the candidate declines to give an independent route to the anchor, or refuses documentation, that refusal is itself a written fact; rescore with it and the case usually lands at Withdraw.

Drift evidence against role access

Tier 2 accessTier 0 or 1 access
Step Up
Run the tier 2 check and withhold access until it clears
Pause, then likely Withdraw
Freeze before offer, shipping, or access; withdraw on a second 2
Proceed
Log the flag and run the standard tier check
Pause and re-tie
Freeze the next round until the anchor confirms or contradicts
No dimension at 2One dimension at 2
A coarse first read for a candidate with a live flag; the written rules decide every edge case.

Each verdict carries a next step and a gate it must clear before:

VerdictNext stepMust happen before
ProceedLog flag, run standard tier checkNext round, as scheduled
Step UpRe-tie anchor, add tier checkOffer is extended
PauseFreeze, re-tie, request documentsAny further round, offer, or shipment
WithdrawEnd process, keep log, reportEnd of the same day

Fix these rules before a flag ever appears, and do not let a hiring manager renegotiate them case by case.

28%
of managers knowingly advanced a candidate they suspected of dishonesty
From Checkr's survey of 3,000 US managers; deadline pressure beats suspicion unless the rule was set in advance.

A pre-committed rule removes the discretion that pressure exploits, and it gives you the consistency employment law expects. The wording to the candidate should be just as fixed.

Candidate wording by verdict
STEP UP
Thanks for the time so far. Before offers for this role, every candidate at this stage completes the same identity check. It is quick on your side, and the link follows separately. Prior employment for this role is also confirmed directly with past employers through their main line, so there is nothing you need to arrange for that part.

PAUSE (new email, number, or address)
Thanks for letting me know about the new [email / number / address]. Before I schedule the next step, it needs to be confirmed the same way it is for every candidate. Could you [reply from the address you first used / send documentation showing the new address]? Equipment for this role ships only to the address on your verified ID.

PAUSE (round-to-round gap)
I would like one more short conversation, cameras on, with the same interviewers, to go back over your work on [project]. This happens whenever there is a gap in the interview notes that needs closing.

WITHDRAW
Thank you for your time and interest. The decision has been made not to move forward with your application for this role.

Send the same wording to every candidate who reaches the same verdict; change only the bracketed details.

Keep the Withdraw message neutral. The reasons live in the log, not in the email.

The step-by-step read, from flag to day one

The procedure runs from the moment an interviewer notices something to the candidate's first sign-in, with a named owner at each step. Steps 1 to 3 produce the verdict; steps 4 to 9 enforce it at the gates where a wrong verdict becomes expensive.

From flag to day one

  1. Log the flag the same day
    The interviewer writes an observable description of the signal within about 10 minutes, with no inference about the person's origin. Done is a completed flag log entry with the role tier filled in.
  2. Score the five dimensions
    The recruiter scores anchor, channel, consistency, and logistics from 0 to 2 and reads the tier set at req open. Noise scores 1 at most; a 2 needs a written fact that contradicts another written fact.
  3. Re-tie to the pre-outreach anchor
    Reach the anchor through a channel the candidate did not supply, such as an independently found switchboard or direct verification with the prior employer or school. Record confirms, contradicts, or no answer by your deadline, then rescore and apply the verdict rules.
  4. Apply the tier step-up to everyone
    TA ops runs the check fixed for the role's tier, identically for every candidate at that stage and for every candidate who trips the same logged flag. Done is a log line showing the same check ran across the tier.
  5. Give biometric notice and get consent first
    If face matching or AI video analysis is involved, legal or TA ops gives written notice and gets consent before any capture. Done is a consent record dated before the first capture.
  6. Verify identity before the offer
    Complete document-based identity verification and the anchor re-tie before the offer goes out, at minimum. Done is a cleared verification and a confirmed anchor on the record.
  7. Ship equipment only to the ID address
    IT or HR ships hardware only to the address on the verified ID, and any change requires documentation first. Done is a shipping record that matches the ID or a documented change.
  8. Verify at first sign-in, then grant access
    IT or security re-confirms identity at first sign-in and withholds system access until every check clears. Done is access granted after the final check, not before.
  9. Withdraw and report confirmed drift
    End the process the same day with neutral wording, preserve the log, and report to IC3 or the local FBI field office. Done is a closed record and a filed report.

Notes on the steps where the evidence is thin:

  • Re-tie timing (step 3). The documented methods are verifying prior employment and education directly with the institutions, and calling the company switchboard to ask for the person by name. No public source says how long either takes to return an answer. Set your own deadline per tier, write it in the log, and treat "no answer by the deadline" as keeping the case at Pause, not as a confirmation.
  • Inbound applicants with no anchor. A Refolk search on the candidate's stated name and prior employer can surface a public profile or commit history to anchor against. An anchor found after the candidate entered your process is weaker, because the candidate knew someone would look, so it scores 1 until a re-tie confirms it.
  • ID timing (step 6). Sources disagree. Trusona says verify before the offer, Hireassist says at the first call, and Checkr says checks must continue after the offer. My read: before the offer at minimum, then again at first sign-in, because the person on the other end can rotate.
  • Step-up effort (step 4). Greenhouse describes its verification as taking "seconds" for the candidate. The cost of a consistent check falls on your process, not on the honest candidate.
  • Staffing firms and contractors. The FBI says to verify that third-party staffing firms run robust hiring practices and audit them routinely. A secondary summary of earlier FBI guidance adds: request documentation of the firm's check, and if it cannot be provided, assume no check was done. I found no published standard terms for employer-of-record arrangements, and the FBI does not explicitly assign interviews, shipping, or day-one presence to the hiring company, so write those duties into the contract yourself.

When a staffing firm sits between you and the candidate, you need a named counterpart who owns their onboarding and device logistics, not a generic account inbox.

How to step up without discriminating or creating biometric exposure

Run the same check on every candidate at the same stage and role tier, trigger extra checks only from logged observable flags, and get notice and consent before any face matching. A step-up applied unevenly creates the legal exposure it was meant to prevent.

What the guidance says:

  • Selective checking is evidence of discrimination. The EEOC says asking only people of a certain race about their financial or criminal histories is evidence of discrimination, and that employers should apply the same standards to everyone.
  • National origin includes accent. The EEOC's national origin coverage includes accent, cultural traits, and perceived national background. "Sounded foreign" can never be a logged signal.
  • No extra documents by origin. Under the Immigration and Nationality Act, an employer may not request more or different documents based on national origin.
  • Selection procedures carry recordkeeping duties. Selection procedures fall under the federal Uniform Guidelines (UGESP) adverse-impact and recordkeeping rules. The flag log doubles as that record.
  • False positives are a named harm. A later FBI PSA, which recommends an "identity in depth" approach, also warns of false positives that could harm legitimate employees and filter out qualified applicants.

No one has published an EEOC-endorsed procedure for stepping up checks by signal and role tier. The tier model here is my working standard, not a safe harbour. Have employment counsel review your tiers and triggers before you rely on them.

Biometric consent

One later FBI advisory (I-072325-4-PSA), as described in a verification provider's summary because I could not retrieve the full primary text, recommends capturing candidate images for comparison across meetings. Capturing face geometry collides with state biometric law:

StatuteTriggerRetentionEnforcement
Illinois AIVIAAI analysis of video, Illinois-based roleDestroy within 30 days on requestNo stated penalty
Illinois BIPAFace geometryPurpose met or 3 years, whichever firstPrivate right of action
Texas CUBIBiometric capture1-year capAttorney general only
Washington RCW 19.375Biometric captureNot verifiedAttorney general only

Practical consequences:

  • AIVIA requires notice, an explanation, and consent before the interview, but only for AI analysis of video for Illinois-based positions. A video interview that a human reviews and scores falls outside it entirely.
  • Under BIPA, each scan is a separate claim and the retention schedule must be posted publicly. An automated face match on an Illinois-based role without written consent compounds per scan.
  • Ask any verification provider where ID and selfie images live. Greenhouse, for example, states it does not store the ID or selfie images collected through its CLEAR integration.
  • Whether a human comparing two screenshots counts as biometric capture under these statutes is not established in the sources I have. Treat any automated match as covered and ask counsel about the rest.

How this read goes wrong

Each dimension has a characteristic false positive or false negative, and the false positives fall hardest on honest candidates. These are the failure modes to audit for, each with the check that catches it.

  1. Rented anchor (anchor strength). A sourced profile can itself be rented, since facilitators provide access to professional networking accounts. Check: was the anchor's history active before the role opened, and is it reachable outside the candidate's contact details?
  2. Looped channel (channel continuity). Calling a "prior employer" on a number the candidate supplied reaches the fraud's own people. Check: find the switchboard number independently, every time.
  3. Delivery scored as drift (cross-round consistency). A new headset, different lighting, or an AI-polished answer gets scored as a different person. Check: compare stable facts like project details and timeline, never delivery or accent.
  4. Honest mover frozen out (logistics drift). A candidate who genuinely moved gets withdrawn. Check: ask for documentation of the new address, the FBI's own remedy, before any withdrawal.
  5. Selective step-up (role access risk). Checks get stepped up only for candidates with "foreign-sounding" names or accents. Check: a log showing the same stage check for every candidate in the tier.
  6. Unconsented face match (step-up). An automated face match on an Illinois-based role runs without written consent or a posted policy, and under BIPA each scan accrues a separate claim. Check: a consent record dated before the first capture.
  7. Vendor flag treated as verdict. A "Flagged" status, which can mean nothing more than VPN use, gets treated as a failure. Flags outnumber failures about 8 to 1. Check: score the underlying fact, not the badge.
  8. Clean background check treated as closure. Records can be pre-tested, so a clean check does not tie the person in the room to the identity. Check: re-tie to the anchor anyway.

What each check actually ties together

  1. Record
    A background check confirms a clean record exists; it can be pre-tested
  2. Document
    ID verification ties the record to a document; test accounts can be bought
  3. Face
    Consented image comparison ties the document to the person on camera each round
  4. Anchor
    A pre-outreach anchor, re-tied on an unsupplied channel, ties the person to a history they did not shape
Each layer of identity in depth links a different pair of things, and only the anchor was set before the candidate knew anyone was looking.
A background check verifies a record. Only an anchor you found first ties that record to a person.

Who carries the check, and how to keep the read current

On most commercial teams the recruiter carries the identity check, because dedicated verification staff are scarce and concentrated in government work. Plan the process on the assumption that nobody else will catch what the recruiter misses.

In Refolk's index of professional profiles, the numbers are lopsided:

Query in Refolk's indexMarket AMarket BMultiple
Technical recruiter or sourcerUS 21,681UK 72629.9x
Background screening or investigator titlesUS 2,764India 25610.8x
US recruiters per US screening specialist21,6812,7647.8 : 1
7.8 : 1
US technical recruiters or sourcers per US background-screening specialist
From Refolk's index; "Background Investigator" dominates the screening titles in both the US and India samples.

The concentration makes it worse. In a 25-profile sample from Refolk's index, US background-investigator employers include Peraton, the US Army, and the US Air Force. In the India sample, a single screening firm accounts for 10 of the 25 profiles. Commercial hiring teams rarely have a verification specialist down the hall. The top-employer patterns come from 25-profile samples only, so treat them as directional and check your own market before relying on them.

To keep the read current:

  • Re-check the advisories. The FBI publishes these warnings as PSAs on IC3, and the State Department issues separate alerts. Read new ones for added signals and add them to the drift list only when they describe an observable, origin-neutral fact.
  • Re-check the statutes. Biometric law changes by state. Confirm the retention and enforcement columns above with counsel before you add any face matching, and fill in the Washington retention cell, which is not verified here.
  • Calibrate the graders. Have two recruiters score the same logged cases blind. If they disagree on a 2, one of them scored an impression instead of a written fact; fix that before the next live case.
  • Audit the log for consistency. Pull every candidate in each tier and confirm the same check ran at the same stage. That record is your answer to both a fairness challenge and a missed fraud.

Before you act on the verdict

  • The flag log names an observable signal and contains nothing about origin, accent, or name.
  • Every 2 in the score cites a written fact that contradicts another written fact.
  • The role tier was set when the requisition opened, not after the flag.
  • The re-tie used a channel you found yourself, not one the candidate supplied.
  • The same tier check ran for every candidate at this stage, and the log shows it.
  • Any face matching has a consent record dated before the first capture.
  • Equipment ships only to the ID address or a documented change.
  • System access stays withheld until every check has cleared.
  • The candidate message matches the fixed wording for this verdict.
  • A Withdraw verdict has a report filed with IC3 or the FBI field office.

Questions practitioners ask

How can I tell if a different person is interviewing in a later round?

Compare stable facts, not delivery. Go back to your notes from the earlier round and test the specifics the candidate gave: project details, timeline, team size, systems named, reasons for leaving. A changed voice quality, camera, or lighting is noise. A candidate who cannot recall work they described in detail last round, or background noise that sounds like a room of others doing similar work, is a written fact you can score as drift.

What should I do if a candidate asks to ship the laptop to a new address after the offer?

Pause and ask for documentation of the new address rather than withdrawing. Honest people move between offer and start, and the FBI's own remedy for an address change is more documentation. Ship only to the address on the verified ID or a documented change. If the candidate refuses documentation, or asks you to ship to a third party, that refusal is a written fact and the case usually rescores to Withdraw.

Is it discriminatory to run extra identity checks on only some candidates?

It can be. The EEOC treats applying checks only to people of a certain group as evidence of discrimination, and national origin covers accent and perceived background. Run the same baseline check for every candidate at the same stage and role tier, and trigger extra checks only from logged, observable flags, applied the same way to everyone who trips that flag. Keep the log; it is your evidence of consistency.

Does a clean background check prove the candidate is who they say they are?

No. A background check verifies a record, not the person in the room. The FBI reports that fraudulent workers buy accounts from identity verification providers to pre-test whether their identities will pass, and the KnowBe4 hire passed four video interviews, a background check, and references. Tie the person to an anchor you found before outreach, through a channel the candidate did not supply.

Do I need consent to compare a candidate's face across interviews?

If the comparison is automated face matching, assume yes. Illinois BIPA treats each scan as a separate claim and requires a publicly posted retention schedule; Texas and Washington also require consent. Illinois AIVIA requires notice and consent for AI analysis of video for Illinois-based roles. Whether a human comparing screenshots counts is not established in the sources here, so ask counsel before relying on it.

Should identity be verified at the first call or before the offer?

Published guidance disagrees: one source says before the offer, one says at the first call, and one says checks must continue after the offer. My working rule is to complete verification before the offer at minimum and repeat the tie at first sign-in, because the State Department reports that the person you deal with can change by time of day.

Try it on the search you came here for

Stop building boolean strings. Just describe the person.

Type one sentence. I plan the search, read GitHub, public LinkedIn and Crunchbase records, and the open web as it is right now, and hand back a ranked list with the reason next to every name.

  1. 01Describe them

    One plain sentence. Role, city, stack, stage, whatever matters to you.

  2. 02I read the web live

    GitHub, public LinkedIn and Crunchbase records, the open web. Not a database that went stale last quarter.

  3. 03You read the shortlist

    Ranked, with the reasoning under every name. Open a profile, ask a follow-up, narrow it down.

  • No boolean, no filters, no seat to buy. One box.
  • Read at search time, so a profile updated yesterday counts today.
  • Every step visible as it runs, every name with its reason.

500 free credits on sign-up. No card, no demo call. See real searches.

Read next