One headline metric: Time-to-Productivity (TTP) — days from start until a receiver handles chilled loads solo at quality parity with veterans.
Baseline: 41 days (median of last 3 cohorts, reconstructed retroactively from scan-log + error data — see attribution note). Target: 26 days.
Everything else in this framework exists to (a) move TTP earlier, (b) catch the people TTP would otherwise lose, or (c) keep the number honest.
1. Baseline definition — what “41 days” actually means
- Population: last 3 cohorts of chilled-receiving receivers (n = 19 starters, 14 reached parity).
- Parity definition: two consecutive shifts with error rate inside the veteran band (≤2.1 errors/100 picked lines), confirmed independently by shift lead + L&D from dashboard data.
- Measurement: TTP day back-calculated from scan-log timestamps + error records; where records were ambiguous, the later date was used (conservative).
- Known baseline flaws: 5 starters quit before parity and are excluded from the median — which flatters the baseline. The honest read is “41 days for survivors, and 26% of starters never got there.” Both numbers are published together, always.
2. Instruments per checkpoint
| Checkpoint | Instrument | Pass / target | What a miss triggers |
|---|
| d5 | Orientation check (name 4 gates) + diagnostic module attempt | diagnostic only — measures program, not person | Explainer/buddy adjustment for next cohort |
| d20 | Flow audit: full load, ≤1 prompted intervention; system-entry lag scored | pass → solo-with-net on schedule | Buddy stays close one more week; card-deck reps doubled |
| d30 | Skills gate: practical 4-gate run + 3 scenario questions (rubric-scored, shift lead + L&D) | cohort pass-rate ≥80% | Cohort pass <80% → buddy-hours +50% for that cohort; individual miss → dated recovery plan |
| d35 | Solo error-rate vs. veteran band | ≤2.1 errors/100 lines | Individual recovery plan, named owner + date |
| d41 | TTP day itself: two consecutive parity shifts, dual sign-off | target d26, ceiling d41 | Anyone past d41 without parity → escalation to L&D owner + ops manager |
| d60 | Parity re-check + rotation status | parity confirmed, rotation underway | Individual recovery plan if parity lost |
| d90 | Retention interview + cohort dashboard | retention + TTP median reported with confounds | Program review; failure-mode register updated |
3. Cohort dashboard (mock — what management sees monthly)
| Cohort | Starters | Reached parity | Median TTP | d30 gate pass | Week-2 retention | d90 retention |
|---|
| C-24 (pre-program, reconstructed) | 7 | 5 | 41 d | n/a | 71% | 57% |
| C-25 (pre-program, reconstructed) | 6 | 5 | 41 d | n/a | 83% | 67% |
| C-26 (pre-program, reconstructed) | 6 | 4 | 43 d | n/a | 67% | 50% |
| C-27 (first Academy cohort) | 7 | pending | pending | pending | measured live | pending |
Pre-program rows are retrospective reconstructions and are labeled as such. The first live cohort row is where claims begin — nothing above the line counts as program evidence.
4. Intervention triggers (pre-committed, in writing)
- d30 cohort gate pass-rate <80% → buddy-hours +50% for that cohort, funded from the L&D budget, not absorbed by shifts.
- Any individual >5 days past the parity ceiling (d46) → named recovery plan with owner and date; escalation to ops manager if d55 passes without parity.
- Week-2 quit → exit interview within 5 working days; findings entered in the failure-mode register and reviewed at the next program review.
- Silent-error signal (scan-log vs. error-log mismatch twice in one week for one person) → buddy runs the recovery ritual re-brief; this is a coaching trigger, never a disciplinary one.
- Cohort TTP median misses target by >7 days → full program review before the next cohort starts; no silent continuation.
5. Honest attribution note
- TTP moves with season, cohort size, dock staffing and veteran availability — a winter cohort with two veterans on sick leave is not comparable to a summer cohort, and the dashboard says so next to every number.
- The program claims association with TTP improvement for its first two live cohorts, not causation. A clean causal claim would need matched cohorts across a full year; that is stated as the evaluation goal, not as a current finding.
- Kirkpatrick framing: L1/L2 (reaction, gate pass-rates) are program-owned; L3 (solo error-rate behavior) is jointly owned with ops; L4 (incident reduction, throughput) is explicitly not claimed in year 1 — the sample is too small and the confounds too large.
- The 41-day baseline is a reconstruction. It is the best number available and it is labeled as such; the first live cohort is the real starting gun.
Fictional disclosure: VanDijk Logistics, cohort figures and thresholds are a fictionalized composite built from realistic logistics onboarding practice. Framework drafted with AI assistance; metric definitions, triggers and final copy are human-owned.