The Failure Prediction Engine Framework monitors five leading indicators across every active mandate and calculates a failure probability score updated daily. When probability exceeds 65%, the engine surfaces a recovery recommendation. When it exceeds 85%, it triggers an automatic recovery sequence. The framework is based on the observation that 68% of VP searches that stall past week 10 showed at least three predictive signals by week 4 — signals that were visible but unmonitored.
The Core Insight
Search failure is not sudden. It accumulates. Response rates decay gradually. Pipeline stages slow before they stop. Hiring managers become less responsive before they go silent. Each signal is individually dismissible — but in combination, they form a pattern that predicts failure with high reliability if detected early enough to act.
The Failure Prediction Engine framework formalises this pattern detection. Rather than waiting for a search to visibly stall, it tracks signal trajectories and calculates a composite failure probability — giving recruiting leaders a 3–6 week window to intervene before the mandate reaches the point of collapse.
"68% of VP searches that stall past week 10 had at least three predictive failure signals present by week 4. The signals were there. No one was monitoring them."
The Five Predictive Signals
| Signal | Early Warning (Wk 2–4) | High Risk (Wk 4–6) | Failure Zone (Wk 6+) | Prediction Weight |
|---|---|---|---|---|
| Response Decay | Rate drops 25% WoW | Rate drops 50% from baseline | Under 8% reply rate | 30% |
| Velocity Stall | Stage avg +3 days | Stage avg +7 days | Stage avg +14 days | 25% |
| Shortlist Rejection | 2 consecutive declines | 3+ declines, same reason | HM stops responding to shortlists | 20% |
| HM Silence | 48–72hr response lag | 72–96hr lag | No response to 2+ submissions | 15% |
| Recruiter Overload | 5 active mandates | 6 active mandates | 7+ mandates, zero capacity | 10% |
Frequently Asked Questions
How far in advance can failure be predicted?
With full telemetry, the Failure Prediction Engine detects failure probability above 65% an average of 4.2 weeks before the search visibly stalls. This window is sufficient to execute a recovery playbook, re-brief the recruiter, adjust the outreach strategy, and re-engage the hiring manager — all before the candidate pool is exhausted.
What triggers an automatic recovery sequence?
When failure probability exceeds 85% and no manual intervention has been recorded in the system for 72 hours, the engine triggers the highest-priority recovery playbook for the dominant failure signal. The action is logged, timestamped, and surfaced to the recruiting manager for review.
Is this different from the Hiring Health Score?
Yes. The Hiring Health Score measures the current state of a mandate. The Failure Prediction Engine measures trajectory — the rate and direction of change across signals — to calculate future state probability. A mandate can have a Health Score of 60 (degrading but not critical) while carrying an 80% failure probability if its signals are all declining simultaneously.