Rollout
From pilot to operation: what must exist first, in which stages to roll out, and where a rollout fails.
What must exist before the first area
- Ownership. A named person per task, not a department.
- Review step. Who checks what, to what depth, and is the time planned in?
- Reporting route. How does someone report that something went wrong, without penalty?
- Training. Tied to this task and this tool, not general.
- Logging. Set up and tested, not planned.
The stages
- 01
One area, four weeks
The same measurement frame as the pilot. The pilot's figures are the expectation.
- 02
Three areas, six weeks
Here you find out what in the pilot was down to the team and what to the task.
- 03
Fully, with handover to operations
Only once ownership, monitoring and reporting routes are in place.
What is different in operation
| Point | In the pilot | In operation |
|---|---|---|
| Users | volunteers, motivated | everyone, mixed attitudes |
| Cases | selected | including the edge cases |
| Review | complete | sampled |
| Attention | high | normal |
| Errors | get noticed | get noticed where review happens |
The last row is the core: what attention caught in the pilot is caught in operation only by the built-in review step.
Monitoring in operation
- Correction rate per area and week. A rise is the first signal.
- Refusal rate. A drop means more guessing.
- Distribution of inputs against the pilot. A shift invalidates the measured quality.
- Share of tasks that come back after the result.
- Usage over time. A drop after eight weeks means the task was not hit.
Rolling back
A rollout needs the same route back that it had forward:
- 01
Trigger named
Which figure, which threshold, who decides.
- 02
Fallback rehearsed
Not described but carried out once, with a measured duration.
- 03
Communication prepared
Who learns what, in what order.
- 04
Cause documented
Not only fixed. Otherwise the same mistake recurs on the next task.