Automation strategy:
| Field | Entry |
|---|---|
| Author | |
| Track | Band 6 quality assurance / Band 7 test engineering / Band 7 test management |
| Area covered | |
| Version and date | |
| Reviewed by |
1. Summary
In three or four sentences: what this strategy changes, why, and what it will achieve in 12 months.
2. Context
- The products and services in scope, and who uses them.
- The current state: how many manual regression cases, how long a regression run takes, which automated tests exist at each layer (from the Module 7 analyses).
- Clinical and operational risks that matter most (from the hazard log).
- Constraints: environments, data, tools, people, regulation (for example IEC 62304).
3. Goals
| Goal | Measure | Baseline | Target | By when |
|---|---|---|---|---|
Link each measure to the metrics proposal.
4. Principles
Start from these and adapt:
- Test by risk, including clinical risk.
- Push tests down the pyramid: unit, then API and integration, then a small number of browser tests for critical user journeys.
- Every automated test makes real assertions, and every suite has a spec that agrees with its code.
- Synthetic data only. No secrets in source control.
- Automated tests trace to hazards and safety controls, and their results are kept as safety case evidence.
- The whole team owns quality.
5. What to automate, and at which layer
| Area or journey | Risk | Layer | Keep manual? | Why |
|---|---|---|---|---|
| Unit / API / Browser |
6. What stays manual
Exploratory, usability, accessibility judgement, and acceptance by clinical users. Say how manual and automated testing work together.
7. Tools and frameworks
Tools (default: JavaScript with Selenium and Mocha, the organisation's CI service), shared helpers or frameworks (see Frameworks and non-functional testing), and how existing suites in other tools are handled.
8. Pipelines
Which suites run on every pull request, which run nightly, and which block a merge or release. How flaky tests are quarantined, with an owner and a deadline.
9. Test data and environments
How synthetic data is generated and maintained, including rare and edge clinical cases. Which environments run which suites.
10. Traceability and evidence
How tests link to the hazard log, and how results are kept as evidence for the clinical safety case and, where relevant, medical device records.
11. People and skills
Who writes, reviews, and maintains automated tests. Skills needed and how they are built (link to this programme and to Leading teams through automation adoption for Band 7 test management).
12. Roadmap
| Quarter | Milestones |
|---|---|
13. Risks
| Risk | Mitigation | Owner |
|---|---|---|
14. Review
When and how this strategy will be reviewed, and by whom.
View this document on GitHub materials/tracks/automation-strategy-and-metrics/automation-strategy-template.md