Bring a small, trustworthy set of production records that illustrates the problem instead of overwhelming the review with every report the system can export.
Start With the Decision
If the question is whether to add a machine, gather representative demand, routings, current capacity, setup frequency, queues, outsourcing, and downtime. If the question is setup loss, collect actual changeover observations and readiness conditions.
The decision determines which data matters. A universal consulting data pack usually creates work without clarity.
Preserve Definitions
Write down what each measure includes. Does downtime include planned maintenance? Does setup end at the first piece or first accepted piece? Does completed output mean machined, inspected, packed, or shipped?
A number without its definition can create false comparisons across shifts, machines, or systems.
Show Exceptions, Not Just Averages
Include examples of the orders, setups, failures, or queues that drove the request. Averages can hide infrequent events that cause the largest customer impact.
Use a Data Confidence Label
| Label | Use |
|---|---|
| Recorded | System or log data with a known definition and time period. |
| Observed | A direct sample tied to a specific job and operating condition. |
| Estimated | Useful directional input that still needs validation. |
| Unknown | A gap that may become part of the measurement plan. |
Labeling confidence is more useful than blending unlike numbers into a precise-looking average.