The market for shop floor software is crowded, and most products look convincing in a demonstration. The choice becomes easier when it follows the use cases instead of the feature lists, and the rollout becomes faster when the pilot produces a standard.
Requirements first
Write the requirements from the use cases of the first two waves: what must the system do for machine data, reasons, boards, Andon? Add the must-haves that knock out a vendor, for example data export in an open format or support for your controls.
Then score with weights that add up to 100:
| Criterion | Weight |
|---|---|
| Fit to the use cases | 25 |
| Ease of use on the shop floor | 20 |
| Integration with ERP and machines | 20 |
| Total cost over five years | 15 |
| Vendor stability and support | 10 |
| Security and operation | 10 |
In one plant, three vendors scored 77, 75 and 74 percent. A result that close does not decide anything.
Test with real users
The plant ran a two-week test with the two best vendors on two machines, with real operators and the real daily meeting. The operators chose the vendor where entering a stop reason took two touches instead of five. On the shop floor, that difference decides whether the data is complete.
Ask every vendor the same questions: how do we export all our data, which controls do you connect without custom work, how are updates tested, what happens when your service or our network is down, and what is the total cost over five years for our roadmap?
A pilot with exit criteria
A pilot takes three to four months and ends with clear criteria: data validated at the machine, the daily meeting running on the board, the paper report removed and the benefit measured against the baseline.
Measure adoption month by month. In the example, stops with a reason rose from 54 percent in the first month to 93 percent after six months. OEE rose from 61 to 68 percent.
Write the standard, then roll out
Before the second area starts, write the rollout kit: OEE definition, reason code catalog, board layout, meeting standard, training plan and go-live checklist. Each wave of two to three months then uses the same kit instead of starting from zero.
Count the benefit conservatively
For twelve machines, seven OEE points free about 4,350 machine hours per year. Counting only half of them at EUR 85 per hour gives about EUR 185,000. Shift reports no longer written add about EUR 62,000. Against an investment of EUR 234,000 in the first year, the payback is about eleven months of operation.
Measure it every month with a named owner. A business case that is never checked is a promise, not a result.
The base for what comes next
Clean machine states, reason codes and master data are also the foundation for analytics and AI on the shop floor. A plant that built that base well can add new use cases quickly. A plant that skipped it will find that no algorithm can fix data nobody trusts.