Why it matters
Automation is a means to an operational result. A business case tied to a specific, measured constraint can be tested before and after deployment. A case built on general expectations cannot.
This question also sets the standard for the other seven. Facility, network, and data requirements all follow from what the automation is expected to achieve.
What can go wrong
- The automated step is not the step that limits output, so total throughput does not change.
- Savings are modeled on labor figures that leave out exception handling, maintenance, and system support.
- Success measures are written after vendor selection and shaped by what the chosen equipment does well.
- Design volumes are taken from a single peak period or a single year and do not reflect the real order profile.
What evidence to examine
| Evidence | What it can establish |
|---|---|
| Documented business objectives and success measures | Whether the project has a testable definition of success |
| Throughput and constraint analysis by process step | Whether the targeted step is the one limiting output |
| Labor data by function: headcount, turnover, overtime | The size and location of the labor problem |
| Order and volume profiles for 12 to 24 months, including seasonality | Whether design volumes reflect actual demand |
| Investment model with assumptions listed | Which assumptions drive the return, and how sensitive it is to each |
| Record of alternatives considered | Whether process or layout changes were evaluated first |
Warning signs
- The project is described by the equipment (“we are getting AMRs”) and not by the outcome.
- Different executives give different answers when asked what success looks like.
- The return calculation came from a vendor proposal and has not been rebuilt internally.
- No baseline measurements exist for the process being automated.
What to do next
- Write the problem statement in one sentence, with a number and a unit.
- Record the current baseline for that number, and how it is measured.
- List the assumptions in the investment model and mark each as verified or unverified.