Do not continue.
Benefit is unsupported, burden is too high, or a hard requirement failed.
Pilot + transition
A pilot is a controlled test—not an unannounced service launch.
A selected concept reaches real-world testing only after the responsible entity establishes the agreement, host, operator, scope, access, data, fallback, stop authority, and evaluation needed for that specific pilot.
Pilot readiness
Twelve items that must be named.
A beautiful prototype can still be unready for field use.
| Readiness item | Required content |
|---|---|
| Site + dates | Where, when, duration, operating hours, and who is in scope. |
| Underlying service owner | The entity responsible for the real service and decisions that exist outside the pilot. |
| Agreement / funding | The instrument that authorizes work, payments, responsibilities, and risk. |
| Baseline | What happens today and what the test is intended to change. |
| Accessibility | Equivalent access, communication, physical/digital access, and support during the test. |
| Notice / consent / permissions | What participants need to know and any consent/permission required for the activity. |
| Minimum data | What is collected, purpose, access, security, retention, deletion/correction, and incident route. |
| Fallback | How the underlying service continues when the experimental approach fails or is declined. |
| Stop authority | Who can pause or terminate the test immediately. |
| Evaluation | Primary outcome, counter-metrics, method, evaluator, and limitations. |
| Concern / incident route | How access, privacy, safety, fairness, or service-loss concerns are escalated. |
| Exit plan | How temporary systems, data, equipment, staffing, and public communication are closed or transitioned. |
After the test
Five possible decisions.
The program should be able to publish a negative or mixed result without converting it into a success story.
Change the design.
The basic approach may remain useful, but field conditions exposed problems.
Run another bounded test.
The revised intervention or weak first evidence justifies another controlled test.
Test transfer.
Try another setting/population/region before claiming broader usefulness.
Begin a separate implementation decision.
Procurement, grant, policy, training, licensing, program adoption, or another mechanism still needs its own authority and resources.