IntermediateQA Engineer

Design a regression testing strategy for a feature

Choose any feature in an application you can access. Map all its user flows, classify each by business risk and change frequency, define which flows belong in smoke, full regression, and exploratory testing, and specify exactly which tests should be automated, which manual, and how each layer fits into the CI/CD pipeline.

Why this matters

Regression testing without a strategy becomes an ever-growing suite that slows CI and still misses important bugs. A regression strategy is the quality contract between QA and the team; it defines what gets tested before every release, what gets spot-checked, and what is intentionally left to exploratory sessions. Building one forces you to confront the real trade-off between confidence and speed.

Before you start

Step-by-step guide

  1. 1

    List all user flows for the feature

    Write out every end-to-end user flow. A flow starts at an entry point and ends at a measurable outcome (data saved, payment processed, notification sent). Aim for 10-20 flows; if you have fewer, you are missing edge cases.

  2. 2

    Score each flow by risk

    For each flow, score two dimensions: business impact if it breaks (1-5) and change frequency (how often this code area is touched, 1-5). Multiply them for a risk score. Sort descending. The top 20% are your regression must-haves.

  3. 3

    Define your three test layers

    Layer 1; Smoke suite: the 5-10 flows that must work for the app to be usable. Run on every commit. Layer 2; Full regression: all flows above your risk threshold. Run before every release. Layer 3; Exploratory: low-risk flows and unusual combinations. Run quarterly or after large refactors.

  4. 4

    Assign each flow to a test type

    For each flow in Layers 1 and 2, decide: automated end-to-end (Playwright), automated API test, manual scripted, or exploratory. Automate flows that are stable, high-risk, and deterministic. Keep manual flows that require visual judgement or have rapidly changing UI.

  5. 5

    Estimate the suite cost

    For your Layer 1 and 2 automation, estimate: number of tests, average test duration, total CI time. If Layer 2 takes more than 15 minutes, it will be skipped. Adjust by parallelising, reducing scope, or moving borderline tests to Layer 3.

  6. 6

    Write the strategy document

    Produce a one-page document: feature scope, risk scoring rationale, three layers with test counts and run frequency, automation boundary decisions, and known gaps intentionally left in Layer 3. This document is what the team refers to when someone asks why a bug was not caught in regression.

Relevant Axiom pages

What to do next

Back to Practice Lab