They picked one number
Every successful pilot named a single metric — first-pass yield on line 3, metal-particle escapes, formation hours per cell — and refused to add a second until the first moved.
- One metric, one line, one quarter
- Baseline agreed by finance before shadow mode ends
- Expansion decided by the number, not by enthusiasm
Pilot designMetric first-pass yield, line 3agreed
Baseline window 6 weeks shadowaudited
Decision gate +3 pts to expandmet
They put quality in the room early
Sites that involved quality and EHS from week one armed control envelopes months earlier than sites that treated it as an IT project.
- Quality owns envelope approval from the start
- Audit trail reviewed before, not after, autonomy
- EHS sign-off on fail-safe behaviour up front
Approval pathQuality engaged week 1on time
Envelope approved week 11faster
EHS sign-off week 9complete
They let engineers correct the model
The plants that improved fastest treated the review console as a training loop, not a chore. Corrections compounded into chemistry-specific accuracy within weeks.
- Named engineers own the proposal queue per shift
- Rejections are captured with a reason code
- Accuracy scoreboard published on the floor
Correction loopProposals reviewed 1,284 / shiftactive
Reason-coded rejections 97%healthy
Accuracy vs engineer +4.1 pts after 8 weeksimproved