One interview, four artifacts, one wording standard · print lives at the tank, digital lives on the tablet
Artifact 1 of 3 · print · A4 laminated · tank wall
Glance-level by constraint, not preference: one-handed, standing, noisy, two languages. Every step has a name and a number — the wording standard the founder asked for. The module teaches the same four gates in the same order and the same words.
Zuivelhoeve De Zwaan · versie 1.0 · NL onder / EN below
Meet zelf met de sonde. Noteer de wachttijd op het erf. Geen overgeschreven cijfers — ooit.
Controleer de cyclusrecord op "voltooid" — niet het lampje. Zeg het hardop voordat je koppelt.
Volg de leiding met je hand. Allergen-CIP-record checken bij elke wissel. Kapot etiket = geen waarde.
Twee vragen ja? Vrijgeven én noteren: tijden, metingen, beide antwoorden, jouw naam.
Was de temperatuuroverschrijding korter dan 30 minuten? · Zakt de kern binnen 20 min op de koelkaak onder 8 °C? — Beide ja → vrijgeven met notitie. Eén nee → houd vast en escaleer.
Zelfde woorden als de module en de KB — één bron. / Same words as the module and the KB — one source.
Trace: steps 1–4 map to interview passages 3, 1–2, and 5–6; the two release questions are the founder's protocol verbatim.
Artifact 2 of 3 · tablet KB · break room
Audience: new operators, week 1–6. Owner: shift lead. Review cadence: after every audit cycle. Lifecycle: drafted from the interview → verified with the founder → owned by the shift leads — the write/verify/own loop the whole kit follows.
A release decision that passes the two protocol questions still doesn't exist for the organisation until it is written down. The note is not admin after the decision — the note is the decision, made auditable. Times, readings, both answers, your name. Trend analysis, supplier conversations and the next person's judgement all run on those notes. The batch you released correctly but didn't record is indistinguishable, later, from the batch nobody checked.
Related: Tank Line Gates (module, Gate 4) · Filling in the log (planned article 2) · floor manual step 4.
Trace: interview passage 6 — “the note is the part everyone skips.”
Predicted questions · from the interview's own hesitations
Artifact 3 of 3 · usability & transfer plan · planned (fictional) — not yet run
Written before the build was called done, because the interview constraint (“survive a glance, one-handed, in boots”) is only provable in use. Status: planned protocol for a concept build — no sessions have been run.
| Step | Task (said aloud) | Watch for | Pass signal |
|---|---|---|---|
| 1 · Warm-up (5 min) | Learner reads the manual card at the tank, holds it one-handed. | Do they find the release box without help? | Found in under 20 s, unaided. |
| 2 · Module run (8 min) | Learner runs the four gates on the tablet, headphones optional. | Gate-2 misses (record vs. lamp); read-aloud skipped? | ≥ 7/8 with best-move reasoning on gate 4. |
| 3 · Transfer probe (3 min) | “New batch, core hit 8.4 °C for 20 minutes — go.” | Do they ask the two questions before deciding? | Both questions asked, then the note named. |
| 4 · Source test (2 min) | “The manual says X, the module says Y — which wins?” | Do they reach for the shift lead / single-source rule? | Answers “same source — flag it,” not a guess. |
Every row maps to a learning objective of the module; failures route back to the artifact named in the trace, not to “more training.”
Success measures · pre-committed, not field results
| Level | Measure | Target |
|---|---|---|
| L1 · Reaction | “Would you use this card on a real Friday?” (3 testers) | ≥ 2 of 3 yes |
| L2 · Learning | Module score on first run + transfer probe | ≥ 7/8 and both questions asked |
| L3 · Behavior | Log entries with yard-delay notes (weekly sample) | Rising toward 100% of arrivals |
| L4 · Results | Excursion batches released without deviation notes | Trending to zero over two cycles |