Best for
- Projects with either no tests or thousands of worthless ones
- Suites so slow or flaky that nobody trusts them
- Deciding unit vs integration vs end-to-end like an adult
What you give it
- The codebase, how it fails today, and what breaking would cost you
- Your tolerance: how much suite runtime is worth how much confidence
What you get back
- A risk-ranked test strategy: what to cover, at what level, in what order
- The highest-value missing tests written and passing
- A flakiness and runtime diagnosis with fixes — quarantine, rewrite or delete, with reasons
How it works
- Maps where the risk is: what changes often, what fails expensively, what guards money and data.
- Audits the existing suite: coverage that matters (not the percentage), runtime, flakiness, redundancy.
- Assigns each behaviour to the cheapest level that can truly catch its failure.
- Writes the missing high-value tests and fixes or removes the harmful ones, with reasons recorded.
- Sets the conventions so the strategy survives: naming, structure, what every new feature must ship with.
Example
You: We have 1,800 tests, the suite takes 40 minutes, it fails randomly, and bugs still reach production weekly.
Result: Diagnosis: 60% of runtime in redundant end-to-end tests of logic already unit-tested; the random failures traced to three tests sharing state. Strategy: a risk map of the five modules producing the production bugs, 24 new tests at the right levels, the redundant layer cut — 40 minutes to 9, and the next two production bugs were caught in CI.
Limits — please read
- 100% coverage is not the goal and it will say so; meaningful coverage of real risk is.
- Deleting tests is proposed with evidence, never done silently.
- Some confidence only comes from production monitoring; the strategy says where tests stop.