Key takeaways
- A test strategy starts from context: the product, its users, its risks, the team and its constraints. A strategy copied from another company rarely fits.
- It answers a handful of questions: what matters most, which levels and types of testing, who tests what, how much to automate, where tests run and when we're done.
- Keep it short, shared and alive. One or two pages, agreed by the team, revisited when things change.
Where this picks up
You know the difference between a test strategy, a test plan and a test case. This article is about writing the strategy: the questions to answer, the order to answer them in and the traps to avoid.
For the deeper topic of making risk the backbone of the strategy, see Advanced risk-based testing.
Step 1: Understand the context
A strategy that ignores context is a template, not a strategy. Find out:
- The product: what it does, how it's built (monolith, microservices, mobile, embedded), how often it's released.
- The users: who they are, what they can't live without, what regulators expect.
- The risks: what failures would hurt most. Start with a quick risk analysis.
- The team: skills, size, who's a tester, who's a developer, who owns quality.
- The constraints: budget, deadlines, tools you must use, environments you have.
Step 2: Answer the core questions
| Question | What to decide | Example answer |
|---|---|---|
| What matters most? | Quality priorities, from the risks | Correct payments and data privacy first; speed second |
| Which test levels? | Component, integration, system, acceptance | Developers own unit and integration; testers own system; product owner runs acceptance |
| Which test types? | Functional, performance, security, accessibility… | Performance test before every major release; security scan in every build |
| How much automation? | What to automate, at which layer, with which tools | Unit + API automated in CI; ~30 end-to-end journeys in Playwright; exploratory for new features |
| Which techniques? | How tests are designed | Decision tables for pricing rules; state models for orders; session-based exploration |
| Where do tests run? | Environments and test data | Each pull request gets a temporary environment; anonymised data, never real customer data |
| How do we handle bugs? | Defect process | Severity set by tester, priority at daily triage; critical bugs block release |
| When are we done? | Entry and exit criteria | No open critical bugs; all high risks tested; product owner accepts the rest |
You don't need all of these at the same depth. Spend most words where the risks are.
Step 3: Decide who does what
Unclear ownership is one of the main reasons testing fails. Make it explicit:
- Developers: unit and integration tests, fixing broken builds, testing their own changes.
- Testers: system-level and exploratory testing, test design for risky areas, coaching others on testing.
- Product owner: acceptance criteria, acceptance decisions, risk acceptance.
- Operations / SRE: monitoring, alerts, rollback.
In Agile teams, many of these are shared. Write down who leads each one anyway.
Step 4: Write it short
A strategy that nobody reads doesn't exist. Aim for one or two pages:
- Purpose and scope: which product, which teams.
- Quality priorities and top risks: five to ten bullet points.
- Approach: levels, types, automation, techniques, in a table like the one above.
- Roles: who leads what.
- Environments and data.
- Defect process and release criteria.
- What we deliberately don't do, and why.
Link to details, such as tool guides and test data scripts, instead of copying them in.
Step 5: Keep it alive
Review the strategy when something significant changes: a new architecture, a serious production incident, a new regulation, a change in team. A short review every few months is enough otherwise.
A company had a 40-page test strategy, written by a consultant, based on a standard template. It described every test level and type in detail. Nobody on the teams had read it.
A new test lead replaced it with a two-page version built from interviews with each team. It listed the top eight risks, a table of who tests what and at which layer, and five release criteria.
Within a month, teams were linking to it in pull request discussions: "The strategy says pricing rules need decision table coverage; where's that?"
Lesson: a strategy's value is in being used. Short, specific and agreed beats complete and ignored.
Do
Write "we don't do X, because Y": "We don't run load tests on minor releases, because they don't change the infrastructure; we do on major releases." It turns gaps into decisions.
Don't
Copy a template and fill every section. Generic text like "all testing will be thorough" tells nobody what to do differently.
Check yourself
Your team releases a mobile app every two weeks. Developers write unit tests, but nobody knows who should test on real devices. Which part of the strategy is missing, and what would you write?
Show answer
The roles part (who does what) and the environments part. You might write: "Testers lead device testing; each release is checked on the 5 most used devices from analytics, using a device cloud; developers check one real device before merging UI changes."
