How we verify what
people can actually do
Veryfy's verification engine operates across three layers — automated, semi-automated, and human arbitration. Here's exactly how each layer works, why it's structurally resistant to gaming, and how trust weights are assigned.
The 3-Layer Verification Engine
Not all verification needs a human. Veryfy uses a tiered system — automated checks fire instantly, semi-automated loops bring context from known platforms, and human arbitration handles disputes and edge cases.
Automated Verification
Runs continuously in the background. Checks platform-sourced data, detects consistency anomalies, validates timestamps, and cross-references artifacts against stated outcomes. No human action required.
Semi-Automated Verification
Triggered when a work entry is submitted or a verifier is tagged. Sends structured verification prompts via email or Veryfy's platform to named colleagues, managers, or clients. Responses are parsed and scored for consistency.
Human Dispute Resolution
Activated when a candidate disputes a verification outcome, a verifier contradicts a prior signal, or the system detects a high-conflict scenario. A trained Veryfy reviewer examines all evidence and issues a binding determination.
Entry submitted
Work entry added with artifacts and context
L1 auto-check
Timestamps, metadata, and platform sources validated
Verifier invited
Manager or colleague receives structured prompt
Signal recorded
Response parsed, weighted, added to score
Score updated
All signals composite into LeaDe Score
How Trust Weights Work
Not every verification signal carries equal weight. Veryfy assigns trust scores based on the verifier's relationship proximity, their accountability, and the verifiability of their own identity.
The core principle: the harder a signal is to fake, the higher its weight. A platform-sourced contribution record is structurally harder to falsify than a self-declaration. A manager with a verified LinkedIn record is harder to impersonate than an anonymous peer. Trust weights reflect this hierarchy directly.
Platform Pull — Automated source data
Data pulled directly from integrated platforms (GitHub, Jira, Figma, Notion, Google Workspace). No human in the loop. Structurally non-falsifiable.
Manager Stamp — Direct manager verification
Verified via email domain check and LinkedIn identity. Manager confirms role scope, performance level, and outcome accuracy. Carries second-highest weight due to accountability structure.
Peer Endorsement — Colleague corroboration
Named colleagues with verifiable profiles confirm specific contributions. Signal is weighted by the endorser's own verification density and reputation score on Veryfy.
Self-Declaration — Candidate-stated data
Lowest weight. Triggers flags if not corroborated by higher-tier signals within a decay window. Self-declared entries that remain unverified decay faster than corroborated ones.
A single strong L1 Platform Pull signal (40%) + one Manager Stamp (30%) gives an entry 70% trust coverage — enough to move it to Verified status in the UI. Self-declaration alone never reaches the Verified threshold, regardless of how detailed the entry is.
Built to Be Ungameable
Every verification system faces the same adversarial problem: motivated actors will try to fake it. Veryfy's architecture was designed with specific attack vectors in mind.
Collusion rings
Two friends verify each other's fake work entries, inflating both scores without genuine evidence.
Social graph analysis
Mutual verification patterns between accounts with no shared employment history are flagged. Verifiers with low reputation scores have their endorsements down-weighted.
Identity impersonation
A candidate lists a senior person as their verifier, then provides their own email to intercept the verification request.
Email domain + LinkedIn match
Verifier email must match the employment domain of the stated company. LinkedIn identity check required for Manager-tier signals. Unmatched verifiers are downgraded to Peer tier.
Artifact fabrication
Screenshots, docs, or files uploaded as evidence are staged or edited to misrepresent actual contributions.
Metadata + source validation
Platform-linked artifacts are validated against live API data. Uploaded artifacts are cross-checked against stated dates and verifier confirmations. Discrepancies trigger L3 review.
Temporal consistency checks
Contribution timestamps must align with employment dates. Gaps or anachronisms trigger automated flags before a verifier is even contacted.
Contradiction detection
If a verifier's response contradicts earlier signals for the same entry, the system holds the score pending L3 review rather than averaging conflicting answers.
Score volatility monitoring
Unusually rapid score increases trigger an audit. Scores can't spike; they compound over time. An overnight jump from 42 to 91 is structurally impossible without triggering a review.
Verifier reputation weighting
Verifiers who have consistently endorsed candidates who later failed have their future endorsements discounted. Reputation compounds — in both directions.
Verification Decay
Skills are perishable. A verification from 3 years ago tells an employer less than one from 3 months ago. Veryfy's decay model ensures scores reflect current capability, not historical peaks.
Every verification signal has a freshness score that decays over time. The decay rate is non-linear — slow at first, then accelerating. This models how skills actually become outdated: gradually, then suddenly.
Active
Signal contributed within 90 days. Full weight applied.
100% weightDormant
90–180 days. Weight begins to decay. Score reflects reduced confidence.
60–85% weightStale
180+ days. Signal carries minimal weight. Profile score drops noticeably.
Below 60%Re-verified
New work added or verifier confirms continued relevance. Signal resets.
100% weightWhy decay matters for employers: A candidate with a Veryfy score of 84 today has demonstrated recency. Their score reflects verified capability contributed within the past 90 days — not a career highlight from 2019. Decay is a feature, not a bug.
Decay applies at the signal level, not the profile level. A candidate can maintain a high overall score by consistently adding new work, even if older entries decay. This incentivizes ongoing contribution — not one-time setup.
Verifier Reputation System
A verification is only as reliable as the person giving it. Veryfy tracks verifier behavior over time and weights their future endorsements accordingly.
When a verifier consistently endorses candidates who perform well in their next role — as tracked via follow-up signals from employers using Veryfy — their reputation score rises. Their endorsements carry more weight. The inverse is also true.
| Verifier signal type | Reputation effect | Weight modifier |
|---|---|---|
| Endorsed candidate hired & performing | +Reputation boost. Future endorsements weighted higher. | ↑ Up to +25% |
| Endorsed candidate hired & underperforming | Mild reputation penalty. Pattern tracked across ≥3 instances. | ↓ Up to -15% |
| Frequent cross-endorsements with same network | Collusion flag raised. Endorsements placed under enhanced review. | ↓ Down-weighted |
| High specificity responses (detailed, role-specific) | Treated as higher quality. Boosts signal confidence for the entry. | ↑ +10% confidence |
| Vague or generic responses | Flagged as low-quality. Signal carries reduced weight. | → -5% confidence |
| Consistent corroboration across multiple candidates | Established verifier status. Faster processing and higher baseline weight. | ↑ Trusted verifier |
Verifiers who reach Trusted Verifier status (based on track record) are surfaced to candidates as high-value endorsers. Their verification requests are prioritized and their signals carry the highest weight within the Manager tier.
Privacy by Design
Verification requires access to sensitive career data. Veryfy's architecture was built on the principle that verification and privacy are not in conflict — they require each other.
Candidate-controlled visibility
Candidates choose which entries are visible on their public passport, which require a direct request, and which are entirely private. Verification exists even for private entries — employers see the signal without the detail.
Verifier anonymization option
Verifiers can confirm or deny claims without their name being surfaced to the employer. The verification signal is recorded; the verifier's identity is held by Veryfy under trust escrow.
Minimal data transfer
Employers requesting verification access receive a structured verdict — not raw data. The underlying artifacts, verifier names, and conversation threads stay within Veryfy unless the candidate explicitly consents to disclosure.
GDPR & CCPA compliant
Candidates have the right to export, correct, or delete all data. Verification signals tied to a deleted entry are also removed. Employer access logs are retained for 24 months for audit purposes.
No credential storage
Platform integrations use OAuth with minimal scopes. Veryfy never stores login credentials. Platform access can be revoked by the candidate at any time — revocation does not delete already-pulled signals.
Immutable audit log
Every verification action — request sent, response received, score change, dispute filed — is recorded in an append-only audit log. Candidates and employers can request a full export of their activity.
See it in action
Request access to the Talent Portal or explore what a verified candidate profile looks like.