Every employee below is real, held-out test data — not illustrative. Drill from the aggregate result down to risk tier, down to confusion-matrix category, down to a single employee's risk score, driving factors, and suggested intervention.
Two honesty notes: (1) This prototype excludes gender and marital status from individual-level explanations and intervention logic, even though the underlying model still uses them — a full fairness audit across age, gender, and tenure is a required step before any real deployment, exactly as flagged in the companion decks. (2) These numbers are freshly recomputed live from the uploaded dataset in this session. Headcount totals match the decks exactly (247 stayed / 47 left / 210 not flagged / 84 flagged). The fine-grained confusion-matrix cells differ by exactly one employee each — AUC 0.799 vs. 0.798, recall 59.6% vs. 61.7% in the decks — most likely a minor row-order difference between the dataset copy used here and the original Colab run. Practically identical; disclosed here rather than silently forced to match.
Drill down by risk tier
Based on the model's predicted probability of leaving. Click a tier to filter the table below.
Drill down by outcome category
The same four groups from the decks — what actually happened vs. what the model predicted. Click a category to filter.
Business case calculator
Choose how many people to retain, or how much cost to avoid — see the matching business case. Same math as the HR leadership deck, live and adjustable.
47
$
Addressable ceiling: this signal can flag at most 141 people company-wide before they leave (59.6% recall × 237 actual leavers) — that's the most any intervention could ever act on, not a target.
Save rate (the share of those 141 an actual intervention successfully retains) is an assumption, not a measured result — nobody has run this intervention yet.
Cost avoided uses SHRM's 50–200% of average salary replacement-cost benchmark, applied to $57,445 — this dataset's real average annual pay among employees who left.
Individual employees
Click any row for that employee's risk score, top driving factors, and suggested intervention.