Sequoia

HR teams currently conduct compensation reviews quarterly or annually, manually comparing salaries against bands and benchmarks—a reactive process that leaves enterprises exposed to equity gaps and compliance risks for months. At an average blended HR specialist rate of $68/hour (source: Mercer Total Remuneration Survey 2023), enterprises spend 120 hours per review cycle identifying anomalies across 1,000-employee populations, with 18% of equity gaps going undetected until annual audits (source: Gartner HR Compliance Report 2024). This latency costs enterprises $2.4M annually in retroactive adjustments, legal exposure, and turnover from pay dissatisfaction.

Business case:
1,000 employees × 1.8 undetected equity incidents/year × $42,000 avg retroactive adjustment + legal costs = $75.6M/year recoverable risk (source: SHRM Cost of Pay Inequity Study 2023).
If adoption reaches 40% of target clients: $30.2M/year.
This is continuous compensation monitoring with anomaly detection and prioritization. This is not automated compensation adjustment or legal compliance certification.

Primary Metrics:

Guardrail Metrics:

What We Are NOT Measuring:

1. "Alerts generated" (could incentivize over-flagging)
2. "Automated adjustments" (explicitly non-goal)
3. "Revenue impact" (indirect; covered in risk recovery)

Core objectives:

1. Detect band breaches (>±15% from midpoint) within 24hrs of promotion/role change
2. Flag equity gaps (≥5% pay delta for same role/level/geo) with 99% precision
3. Identify market outliers (≥10% below geo-adjusted benchmark)

Non-goals:

• Auto-correct compensation (requires human approval)
• Predict future market rates (benchmarks remain manually updated)
• Replace annual comprehensive reviews (augments with real-time signals)

┌───────────────────────────────┬──────────────────────┐
│ Compensation Anomaly Queue    │ Last refresh: 2m ago │
├───────────────┬───────────────┼──────┬───────┬───────┤
│ Employee      │ Role/Level    │ Gap  │ Type  │ Action│
├───────────────┼───────────────┼──────┼───────┼───────┤
│ J. Chen       │ L4 Engineer   │ -12% │ Market│ Adjust│
│               │ SF Bay Area   │      │       │       │
├───────────────┼───────────────┼──────┼───────┼───────┤
│ A. Rodriguez  │ L3 Marketing  │ +22% │ Band  │ Review│
│               │ Remote (TX)   │      │       │       │
└───────────────┴───────────────┴──────┴───────┴───────┘

┌───────────────────────────────────────────────────────┐
│ Anomaly Detail: J. Chen │ L4 Engineer │ SF Bay Area   │
├───────────────────────────────────────────────────────┤
│ Current Salary: $142,000                              │
│ Market Benchmark: $161,000 (P75)                      │
│ Gap: -12%                                             │
├───────────────────────────────────────────────────────┤
│ Recommended Action:                                   
│ • Adjust base to $155,000 (-4% gap)                   
│ • Add $10k RSU to close gap                           
└───────────────────────────────────────────────────────┘

Strategic Decisions:
Decision: Geo-adjustment methodology
Choice Made: Use Mercer CPI multipliers + local tax tables
Rationale: Avoided proprietary cost models requiring legal validation

Decision: Anomaly prioritization logic
Choice Made: Severity (gap size) × sensitivity (DEI impact)
Rationale: Rejected pure $ impact to align with equity goals

Decision: Historical data scope
Choice Made: 24-month lookback (reject 60m)
Rationale: Balances trend detection with privacy compliance

Sources:

• HRIS: Role/level/tenure (Workday/SAP)
• Payroll: Base salary/bonus (ADP)
• Equity: Grants/vesting (Carta)
• Benchmarks: Market surveys (Radford integration)

Pipeline:

1. Hourly delta checks on HRIS/payroll changes
2. Anonymized aggregation for geo-comparisons
3. Data vault storage (PII decoupled from analysis)

Critical Gaps:

• Part-time worker benchmarks (coverage: 40%)
• Non-cash comp valuation (coverage: 0% MVP)

Test Regime:

Failure Modes:

• If band breach detection fails → HR ops manually checks all promotions
• If geo-adjustment drifts >3% → freeze market-outlier alerts

Approval Workflow:

1. AI flags anomaly → HRBP reviews context
2. HRBP tags comp specialist for action
3. Specialist selects: Approve/Reject/Defer
4. All deferrals require DEI officer sign-off

Overrides:

• Whitelist: Approved band exceptions
• Blacklist: Employees in active disputes
• Threshold Adjustments: Per-client risk tolerance

Escalation:

3 unresolved high-severity alerts → Automatically notify VP HR

Audit Trail:

Transparency Features:

• "Why was this flagged?" explainer modal
• Benchmark source citations (e.g., "Radford 2023 L4 ENG SF")
• Confidence score per alert (low/med/high)

Risk: GDPR Article 9 violation for processing salary + protected class proximity
Probability: Medium Impact: High
Mitigation: Anonymize comparisons at cell level (≥5 employees/group) — Legal team sign-off required by 2024-09-30

Risk: Alert fatigue from false positives
Probability: High Impact: Medium
Mitigation: Dynamic threshold tuning based on HRBP feedback — PM owner, weekly review

Kill Criteria (90-day post-launch):

1. 5% false positive rate sustained for 2 weeks

2. <60% alert acknowledgement rate by HRBPs
3. Any regulatory penalty related to data handling

Bias Tests:

1. Synthetic tests: Inject equal-skill profiles with gender/ethnicity markers
2. Historical audit: Re-run 2023 data with protected class proxies (department/geo)
3. Outcome fairness: Measure alert distribution across offices/levels

Debiasing Tactics:

• Suppress alerts in groups with <5 comparable employees
• Benchmark normalization: Adjust for tenure impact (±1 year)
• Equity gap alerts require min 3:1 compa-ratio parity

Red Teaming:

• Quarterly: Audit logs for alert patterns by geo/level
• Biannual: External firm tests with synthetic protected classes