Compact Kirkpatrick application for a single live event — demonstrates measurement thinking at workshop scale (TU Delft asks for exactly this evaluation literacy; FareHarbor wants "data-driven impact stories").
L1 · Reaction (exit poll, 3 items, 40 seconds)
| Item | Target |
|---|
| "Worth my time" (1–5) | ≥80% rating 4–5 |
| "I will hold the conversation" (yes/maybe/no) | ≥85% yes |
| One thing to improve (open) | themed into next cohort's tweaks |
L2 · Learning (rubric delta — the interesting one)
- Observer rubric scores captured per round (digital rubric:
03-observer-rubric.html)
- Metric: mean Round-2 − Round-1 delta, target +1.0 (of 25) minimum
- Difficulty confound checked: if Round-2 scenarios rated harder by facilitator panel, note alongside delta (transparent analysis beats cherry-picked wins)
- Item-level read: which rubric line lifts least → becomes the next cohort's clinic focus (SBI-C specificity is the usual suspect)
L3 · Behavior (the commitment card does the work)
- T+1 pocket-card nudge (open-rate tracked)
- T+10 manager pulse: one question, two clicks — "conversation held?" + "any visible change?"
- Target: 70% conversations held within 10 working days
- Skip-level triangulation at T+30 via 2-question engagement micro-survey to team members (anonymous): "I get clear, timely feedback" trend line
L4 · Results (deliberately modest — honesty over inflation)
Single workshop cannot claim engagement-survey movement. Committed statement instead: contribute one measurable input (feedback-quality index item) to the quarterly survey; if 3 consecutive cohorts run and the item trends up ≥0.3, report series-level association, never causation. This restraint is a design feature — evaluators trust modest claims backed by numbers.
Reporting artifacts
- One-page dashboard per cohort (L1 poll · L2 delta · L3 completion) — feeds the L&D quarterly review
- Facilitator retro notes → next version of the run-sheet (versioned v1.x like the kits' other artifacts)