The interesting part of this rollout is not the software. It is that two-thirds of the effort went into deciding which of eleven versions of each procedure was actually right.
The problem was not the documents
Every plant had its procedures. The problem was that every plant had its own copies, in Word, on a shared drive, in folders whose naming conventions had diverged years ago.
The immediate trigger was an audit. Producing evidence of which revision of a procedure was in force on a specific date, and who had been trained on it, took the quality team the better part of three weeks and involved emailing plant managers to ask what they had printed.
The second, quieter problem was the laminated sheets. Nobody could say with confidence that what was posted at a workstation matched what was on the drive.
Consolidation was most of the work
The import itself was straightforward: Word documents with real heading styles converted cleanly, and the ones formatted by hand were flagged in the import report.
The hard part took six weeks and had nothing to do with software. Eleven versions of the same changeover procedure had to be compared to decide which differences were genuine - different equipment, different line layout - and which were accidental, the result of eleven people editing eleven copies over five years.
Roughly a third of the differences were accidental. Those became one shared procedure. The genuine ones became plant variants of a shared base, with the common steps held as snippets.
- 900 documents in, 540 after de-duplication.
- About a third of plant differences turned out to be accidental drift.
- Common safety and quality steps became snippets shared across all variants.
- A shared safety-warning change now takes one edit rather than eleven.
Rollout, and the month it did not work
The first plant took five weeks, mostly spent settling the category structure and the snippet library. Once those were fixed, subsequent plants took about four days each.
Adoption in the first month was poor. Tablets were on the floor, the procedures were current, and operators kept using the laminated sheets - because using the tablet required knowing a procedure existed and going to look for it, while the sheet was already on the machine.
The fix was printed QR codes at each workstation, each resolving to the current approved version of that station's procedure. Usage jumped within a fortnight. The laminated sheets came down over the following month.
What the audit looked like afterwards
The next audit cycle took two days of preparation instead of three weeks.
The specific things that made the difference: the exact text of any procedure on any past date, retrievable with its approval record; acknowledgement reports per person, line and plant, exported as CSV; and a list of overdue reviews that the quality team could act on before the auditor found it.
The quality manager's summary was that the auditor asked the same questions as always, and the answers stopped requiring an archaeology project.
What they would do differently
Two things, in their own assessment. Put the QR codes up in week one rather than month two - the first month of poor adoption was avoidable and cost them goodwill on the floor.
And time-box the consolidation debate. Six weeks of comparing eleven versions was longer than necessary; a rule that any difference nobody could justify in five minutes became the shared version would have saved at least two weeks.