Source Quality Scorecards That Predict Interview-to-Retention

Track source quality by downstream evidence or you will keep paying for volume that converts into rework, fraud exposure, and early attrition. The failure pattern is predictable in a CPO war room: a channel delivers a spike of applicants, recruiting hits SLA pressure, and screening shortcuts emerge. Two weeks later, hiring managers complain about interview quality. Two months later, you are backfilling a role and Legal is asking for documentation you cannot reconstruct. The operational risk is not the bad candidate. The risk is that your process cannot prove how a candidate moved from "sourced" to "approved" across tools, reviewers, and identity states. If legal asked you to prove who approved this candidate, can you retrieve it with timestamps, rubrics, and identity continuity? If it is not logged, it is not defensible. Cost shows up twice. First, cycle-time waste: recruiter hours spent on candidates who never clear an evidence-based bar. Second, mis-hire replacement exposure. SHRM estimates replacement cost can range from 50-200% of annual salary depending on role. When a source inflates low-signal volume, you pay that tax more often and you pay it with less defensibility. Fraud makes this worse because it changes what "quality" means. Checkr reports 31% of hiring managers say they have interviewed a candidate who later turned out to be using a false identity. If your scorecards treat interviews as ground truth without identity gating, you can end up optimizing for the channels that most efficiently deliver unverifiable humans.

The market did not solve this because most stacks are built for handoffs, not continuity. Your ATS can store stages. Your background check vendor can return a pass-fail. Your interview tool can store notes. Your coding tool can store scores. None of them are accountable for end-to-end evidence continuity from source to retention. The failure modes are operational, not philosophical:

Assign ownership explicitly or your scorecards become an argument instead of a control system. Ownership and accountability matrix: - Recruiting Ops owns workflow design, source taxonomy, SLAs, and ATS data hygiene. They are accountable for the instrumented funnel and reconciliation when events fail. - Security owns identity gate policy, access control, audit policy, and step-up verification thresholds for higher-risk roles. They are accountable for evidence pack integrity and reviewer authorization. - Hiring Managers own scoring discipline: rubric completion, calibrated interview decisions, and documented rationale. They are accountable for evidence-based scoring, not "gut feel" notes. - People Analytics owns dashboards, segmentation, and retention joins. They are accountable for time-to-event analytics and confidence levels when data is missing. Sources of truth by object: - Candidate record, stage timestamps, source field: ATS is the system of record. - Identity verification state, integrity flags, and evidence pack artifacts: IntegrityLens evidence pack with ATS write-back references. - Interview rubric scores and notes: ATS-anchored rubric objects (not docs) with reviewer identity and timestamps. - Retention marker (30-60-90 day): HRIS is the system of record, joined back to ATS candidate ID with reconciliation jobs.

Instrument source quality as a controlled workflow where every candidate event is attributable, time-stamped, and joined to outcomes. Recommendation: treat hiring like secure access management. A source is not "good" because it delivers volume. It is good because it reliably delivers verified identities that clear evidence-based scoring and sustain retention markers, without inflating review queues or audit risk. Operating model components (answer-first, then mechanics):

IntegrityLens acts as the ATS-anchored control plane that keeps identity, screening, and assessments in one evidence chain so source quality can be measured downstream, not guessed upstream. Operationally, IntegrityLens enables: - Identity gating with biometric verification (liveness, face match, document authentication) before expensive interview time is allocated. - Fraud prevention signals (deepfake and proxy interview detection, behavioral signals) that can be used to route only high-risk sessions into a manual review queue. - AI screening interviews that run 24/7 with consistent prompts, producing structured outputs that can be joined back to source and rubric expectations. - AI coding assessments across 40+ languages with plagiarism detection and execution telemetry, supporting evidence-based scoring rather than subjective impressions. - Immutable evidence packs with timestamped logs, reviewer notes, and zero-retention biometrics architecture, written back into the ATS for audit-ready provenance.

Do not implement source scorecards in ways that create new integrity liabilities: - Optimizing channels on top-of-funnel volume while leaving identity unverified until late stages. You will reward sources that can generate proxies and deepfakes efficiently. - Allowing rubric-free interviews or unstructured notes for some teams. Inconsistent evaluation criteria makes your source comparisons non-defensible under audit. - Running manual exceptions through email or chat without logging rationale and approver. Shadow workflows are integrity liabilities.