Why factory software fails at adoption, and what the floor needs first
The graveyard of factory software is not full of bad products. It is full of good-enough products that the floor quietly refused, until the system became a formality maintained for reports while the real factory ran on paper and memory. Adoption is not a training problem. It is a design and sequencing problem, and it is decided in the first month.
The floor pays the cost, the office gets the benefit
Most implementations invert the value flow: operators and supervisors type all day so managers can see dashboards. Any system arranged that way is running against human nature, and human nature wins. The arrangement that survives gives the person entering data something back at the moment of entry: the supervisor logging hourly output sees their line's pace against target instantly; the storekeeper posting an issue sees the balance update and the pick list shrink. Data capture that answers the capturer is self-sustaining. Capture that only feeds the office is a tax, and taxes get evaded.
Seconds are the currency
A floor entry that takes half a minute happens; one that takes three minutes happens on quiet days only. This is why capture surfaces deserve the same engineering attention as any dashboard: big targets, minimal typing, defaults from context, phone-friendly, tolerant of wet hands and poor light. Every field on a floor form should justify itself: it explains a variance, it feeds traceability, or it goes.
Sequence by trust, not by module list
The classic rollout plan marches through modules alphabetically. The rollout that works marches through trust: start where the system can be visibly right within two weeks. Hourly production capture is often the best opening move, because within days the 07:30 meeting is reading numbers nobody had yesterday, and the floor sees its own work reflected. Each surface that proves trustworthy buys patience for the next. Each surface that ships broken or ignored spends trust you cannot easily reprint.
Parallel running is where systems go to die
Running the old spreadsheet alongside the new system to be safe feels prudent and is usually fatal, because it doubles everyone's work while guaranteeing the two disagree, and the discrepancy is always blamed on the newcomer. The braver pattern is a clean cutover per surface with the old artifact retired the same week, after the new one has proven itself in the two-week trust window.
The signals worth watching
Adoption is measurable. Watch capture latency (how long after an event is it recorded), correction rates, and the one metric that matters most: whether the numbers people quote in meetings come from the system. When a supervisor defends their line by pointing at the screen rather than disputing it, the system has stopped being software and become the factory's shared memory. That is the finish line, and everything in the rollout should be aimed at it.