The test that matters
You cannot tell whether an operational excellence program is going to work on a good shift, when production is ahead and everyone is pleasant. You can only tell on a bad shift — when a jam-up has cost you two hours, the area manager is being paged by the plant director, and the maintenance team is already on another line. Does the system hold, or does everyone revert to whatever they were doing three years ago?
Most OpEx programs fail that test. They were designed for a good day. Here is how we design ours for the bad ones.
1. Tempo the floor can keep
Tiered huddles are standard Lean vocabulary. The mistake most plants make is importing a cadence from another industry. A 90-minute end-of-shift review borrowed from automotive will not survive a 12-hour continuous shift in glass. Our default rhythm:
- T1 (crew, 5 minutes, every 2 hours): what happened, what's next, what's stuck.
- T2 (area, 15 minutes, once per shift): shift scoreboard, top issue, handover.
- T3 (plant, 20 minutes, daily): line-by-line OEE, open escalations, resource conflicts.
- T4 (leadership, 30 minutes, weekly): trend, variance to plan, strategic calls.
2. Standard work written by the people who do the work
The binder on the shelf is not the standard. The standard is what the crew does at 3am with no one watching. The two converge only when the people doing the work wrote the standard, understood why it was written that way, and have a reasonable path to change it when the work changes.
3. Escalation that closes
A huddle that surfaces issues without closing them is a complaints bureau. We track time-from-raise-to-close and publish it at T3. If 85% of escalations do not close within their SLA (usually 48 hours for area, 7 days for plant), the system is theatre.
4. Problem-solving that does not require a PhD
We teach A3 or 4-step problem-solving because the structure is useful, but the goal is to get 50–200 closed problem-solving reports per year with pictures, data and a visible action. Not fifteen ornate ones. The first year of a new OpEx program should celebrate volume as much as quality — the quality comes with repetition.
5. Leader standard work
The single most predictive indicator of a program that will last twelve months: leader standard work. If the plant manager has a daily walking route that takes them past five specific boards in a specific order, and they actually follow it 90% of the time, the system will hold. If they don't, nothing else will save it.
6. Dashboards that reward the right behaviour
Measure the behaviours you want — standard work adherence, escalations closed, problem-solving reports submitted — not just the outcomes. Outcomes lag. Behaviours lead. Without behaviour KPIs, you are steering by looking at the wake.
The rollout we run
- Pilot line, 8 weeks. One line, one crew, one dashboard. Daily coaching. We are on the floor more than in the conference room.
- Plant rollout, 16 weeks. Line-by-line, with internal coaches leading. We coach the coaches.
- Sustain, 6–9 months. Monthly audit and coaching rhythm. We hand over. Bad shifts become the proof point.