You benchmarked the plant. The culture didn't travel.

Here is an uncomfortable observation from two decades of plant walk-throughs: the best-maintained piece of visual management on most sites is the A3 wall nobody uses. Colour-coded. Laminated. Last dated eighteen months ago, the week the delegation came back from benchmarking the sister site. That week, a CHRO touring the same plant gave me a line worth pinning above every transformation office: culture gets translated, never copied. And the news cycle keeps running the invoice for ignoring that – the "DX failure trap" coverage, companies pouring money into transformation programmes, industrial AI included, and roughly five per cent delivering anything measurable. Same disease. Bigger budget.

In a copied lean programme, the most heavily utilised machine on site is the laminator.

What a copy actually copies

Templates, cadences, org charts, terminology. A copy transfers everything except the reasons. The copied QRQC keeps the eight-minute stand-up and quietly loses the authority to stop the line. The escalation timer survives the flight. The promise that escalation is safe does not.

Every working practice encodes invisible assumptions about how a plant really runs: how fast bad news travels upward, what genuinely gets punished, who is allowed to be wrong out loud. QRQC at its best assumes the operator who finds the defect owns the first response – and that saying "I caused this" in front of thirty people is career-safe. Copy the stand-up without the amnesty and you get theatre. Everyone waits for the shift leader to speak, the board fills with someone else's handwriting, and the data improves for exactly as long as the visitor keeps walking past.

Copying imports the ritual and orphans the reason.

Same standard, different dialect

Translation is not softening the standard. The standard stays; the dialect changes. I learned the difference by accident, building the greenfield quality department for a 900-plus employee plant from zero. Nothing to copy – no sister site, no inherited ritual, no laminated history. Every practice had to be re-derived from first principles: why a Q-Wall belongs at the line and not in a meeting room, what makes an operator flag a suspect part in minute one instead of hour four, what the escalation clock is actually for. Slower to build. Also the cleanest starting position I have ever had, because nothing we installed had to out-compete a dead ritual for credibility. The scoreboard agreed: defect costs down 70 per cent, customer satisfaction at 98.

Later, running quality initiatives across a 2,000-plus multi-site workforce, the lesson inverted: one standard, several translations. A three-shift site with a strong works-council tradition needed its escalation path negotiated, not announced. A site still stinging from a customer escalation was hungry – the same 8D discipline could be tightened a notch there without a murmur. The mechanism never changed: stop the line on evidence, escalate on facts, solve at the lowest competent level. The ritual around it was rebuilt site by site, in the local language, fitted to shift patterns and to what each site's history would tolerate.

The proof had come earlier, at WITTE Automotive. The toolkit – QRQC, A3, Q-Wall – was straight out of any textbook, and in textbook form it did little. It began paying, and paid out in substantial, audited failure-cost reduction, only after we translated it: the Q-Wall moved out of the meeting room to where the parts actually flowed, the escalation timer was re-cut to the plant's real decision speeds, and line-stop authority went to named people on shifts rather than job titles on a chart. Identical mechanism. Our own ritual.

The tell, and the audit question

You can tell a copy from a translation by its mortality. A copied practice dies with its champion: the plant manager who went on the benchmark tour moves on, and within a quarter the wall goes quiet, the timer stops being reset, and the meeting shrinks to the people who cannot find an excuse to skip it.

Through FOREAST engagements across automotive, steel and optics – ArcelorMittal, Meopta, AVX Kyocera among them – I settled on one audit question that prices the difference in a single conversation: ask the operator why the rule exists. A translated practice survives the answer. You get the mechanism back in the operator's own words, usually with a local war story attached. A copied practice produces silence, a shrug, or "that's how it's done at corporate". The money follows without mercy: translated practices cut defect costs, copied practices add meeting load. A forty-minute imported daily meeting across four departments, at fully loaded salaries, clears €100,000 a year – with not one defect attached to it.

Playbooks scale pilots; leaders translate them. There is no shortage of playbooks – industrial AI at scale is this season's publishing category – and the five per cent of programmes that deliver will be the ones that did the unglamorous work of re-deriving the culture around the tool: who stops the line, who gets forgiven for raising bad news, how fast a defect travels from the person who found it to the person who can resource the fix. None of that fits on a slide. So run the check on your own transformation: if it arrived as a PDF, it is not deployed. It is laminated.