Who pays for the data
Instrumentation always has a payer, and it is rarely the person who benefits. A field technician logging a fault code is spending their time so that someone three levels up can see a pattern next month. That asymmetry is the reason most field data is late, thin and quietly invented at the end of a shift.
Every second of capture has to be earned. The question to hold through design is not what would be useful to know, but what the person entering it gets back before they go home.
Data that is only useful upstairs will only ever be entered under protest.
Design for the actual conditions
- One hand. The other is holding something, and it matters more than the phone.
- Poor light, gloves, noise, every input that needs precision will be skipped or guessed.
- Intermittent connectivity, so the record has to survive being made offline and reconciled later.
- A shift that is already full. Capture competes with the work, not with idle time.
Give it back inside the shift
The instrumentation that survives is the kind that answers a question the person carrying it already had: where is the part, has this asset failed this way before, who else is nearby. Once the record is worth making for its own sake, quality stops being a compliance problem.
That is also the cheapest data-quality intervention available. No validation rule ever written has done as much for accuracy as making the record useful to the person creating it.
