Draft
Reverse Engineering an Undocumented MES Metric
How to prove whether a production metric is time-per-part or throughput, rather than trusting the name of the field it lives in.
- Reverse engineering
- Manufacturing systems
- Data validation
Draft. An outline of something I worked out in practice and intend to write up properly.
What this will cover
A field called speed can hold seconds per part. Read as throughput, every performance indicator built on it inverts — the machines running behind standard look healthy and the healthy ones look bad. The name of the field is not evidence.
The intended shape of the article:
- Why field names are unreliable in legacy manufacturing systems, and what they are actually describing when they were written
- Establishing units from the data: comparing a derived column against a formula you construct yourself, and checking it reproduces the vendor’s value on every row rather than on average
- Establishing grain: whether a value is instantaneous, per shift, or a job-lifetime aggregate, and why that distinction changes what a dashboard is honestly allowed to claim
- Using the vendor’s own audit tables as evidence of intended workflow
- Controlled tests against live data under transaction rollback
- Writing the finding down so it survives you: evidence, the assumption tested, and what would falsify it
Why it is worth writing
Every one of these checks is cheap. Skipping them produces a system that is confidently wrong, which is considerably worse than one that is obviously broken.