Advice to go in phases is common; what phase one should be is not. Three criteria apply: the effect must be visible, the data must be easy to obtain, and failure must not stop production.
Actually measure where people-hours go.
For each candidate, ask whether the required data is obtainable today.
Phase one should be an area where a problem does not stop production.
Decide numerically what success looks like.
Converting everything at once is not only larger in budget and schedule — it exceeds the amount of change a floor can absorb. When working methods all change together, resistance appears and problems become hard to isolate.
Divided too finely, though, each phase produces no visible effect and funding the next becomes difficult. Phase one has to be big enough that the change is obvious.
Following the order matters more than perfecting each step.
Actually measure where people-hours go. Production logs, result collation, stock counting and drawing distribution are typical candidates. Measure, do not estimate.
For each candidate, ask whether the required data is obtainable today. Available from a machine signal is easy; needing new sensors or new procedures is hard. Phase one comes from the easy side.
Phase one should be an area where a problem does not stop production. Record-and-view functions are better candidates than anything tied directly to control.
Decide numerically what success looks like. It has to be something whose prior value is known — time spent collating results, stock count discrepancy, drawing mis-distribution incidents.
Patterns that recur in implementation discussions.
The place with the most problems is not the place to start. High difficulty lengthens phase one and costs organisational confidence.
Without a baseline measurement, you cannot say afterwards whether it improved.
The purpose of phase one is verification. Trying to cover every exception grows the scope again.
Data structures and identification schemes have to be designed for expansion, or they get rebuilt later.
Confirm these are in place before starting.
The procedure is the same without any product. These shorten the steps.
Questions that come up during evaluation.
Production result collection and floor delivery of drawings and instructions. Manual effort is high, data is comparatively easy, and production-stoppage risk is low.
Long enough to include verification, short enough that the organisation keeps its focus. Stretch it and the original objective blurs.
No. The usual pattern widens the scope of the same system, which is why phase one has to leave the data structure extensible.
Yes. Splitting scope between funded and self-funded phases is common.
Equipment mix and current workflow determine how the procedure is adapted.
Request a site survey