Kit 20 · LMS Pulse · deliverable · designed and built by I Wayan Usada · fictional composite (Meridiaan College and every number are invented)
Dashboards mix three families of numbers. Revised after the build so it matches what ships: the 2D drill and the 3D funnels both read pulse-data.js.
The earlier build said "deltas and drop-offs measure learning". Drop-off is persistence, not learning; that sentence is gone from every artifact.
| # | Cohort | On screen | Right story | Tempting story | Interventions that fit |
|---|---|---|---|---|---|
| 1 | A · Finance systems | 92% completion, 4.6/5, no cliff, −3 pts vs a 71% pre-test, “Confusing!” ×41 | learning problem | healthy, or react to the comment | assessment redesign, content fix |
| 2 | B · Client communication | 38% completion, cliff in week 4 before the workshop, +11 pts, “Too busy” ×29 | persistence problem | learning problem (low completion) | workload fix, manager briefing |
| 3 | C · Compliance refresh | 97%, 4.9/5, +1 pt vs a 94% pre-test | measurement problem | healthy | assessment redesign, do nothing + instrument |
| 4 | Debrief | the three cohorts side by side | merging them into one average |
Step 1: choose one of four stories (learning, persistence, measurement, healthy); the same four for every cohort, shuffled, so length or position never signals the answer (the earlier build offered two options per cohort, one obviously longer). Step 2 unlocks after the diagnosis: pick two of six interventions. In 3D each pick is pinned where it acts: intake, the leaking week, the funnel body or the gauge.
Diagnosis 2 + each fitting intervention 1, per cohort; max 12, pass 10 with all three diagnoses right. Order of the two picks is not scored (the earlier copy said it was, but the code never scored it).
Prefer the source? Download the raw markdown.