Software Testing

Intermediate path · lesson 10 of 19

IntermediateStrategy5 min readPublished November 12, 2026

Test planning and test estimation

How long will testing take? Nobody knows exactly, but you can give a reasoned answer. Here's how to plan the testing of a release and estimate it with techniques that work, including how to communicate uncertainty.

On this page

Key takeaways

  • A test plan turns the strategy into concrete decisions for one release: scope, approach, schedule, people, environments, criteria and risks.
  • Estimate with more than one technique: ratios from past projects, breaking the work down, three-point estimates and team estimation like Wideband Delphi or planning poker.
  • Give estimates as ranges with assumptions, and re-estimate as you learn. A single number hides the uncertainty.

Where this picks up

You know what a test plan contains. This article is about making one for a real release, and the hardest part: answering "how long will testing take?"

Planning, step by step

  1. Read the strategy and the release scope. What's new, what changed, what's risky?
  2. Agree the scope: what's in, what's explicitly out.
  3. Choose the approach per area: which techniques, which levels, what to automate. Let the risk analysis set the depth.
  4. Identify dependencies: builds, environments, test data, other teams, external systems.
  5. Estimate the effort (see below).
  6. Build the schedule around the dependencies, with the highest risks early.
  7. Define entry and exit criteria and get them agreed.
  8. List the project risks and what you'll do if they happen.

What makes testing take longer

The ISTQB syllabus groups the factors that affect test effort. Use them as a checklist when estimating:

FactorPushes the estimate up when…
ProductMany integrations, complex rules, high quality or regulatory demands, vague requirements
ProcessManual deployments, unstable environments, slow tools, heavy documentation
PeopleNew team members, unfamiliar domain, a team spread across time zones
Test resultsMany bugs found, meaning many fix-and-retest rounds

The last one surprises people: the quality of the code affects how long testing takes. Every bug adds investigation, reporting, waiting for the fix and retesting.

Estimation techniques

Ratios and history

Use data from past projects. "On our last three releases, testing took about 40% of development effort." Quick and grounded in your own reality, if the projects are similar.

Breaking the work down

Split the work into small tasks, estimate each and add them up: prepare test data (1 day), checkout test design (2 days), checkout execution (3 days) and so on. Small tasks are easier to estimate than big ones, though people often forget the in-between work like setup, reporting and retesting.

Three-point estimation

For each task, give three numbers:

  • a: optimistic, if everything goes well
  • m: most likely
  • b: pessimistic, if typical problems happen

The weighted estimate is E = (a + 4m + b) ÷ 6, and a rough measure of uncertainty is (b − a) ÷ 6.

Example: testing a new payment provider. a = 4 days, m = 6 days, b = 14 days.

  • E = (4 + 24 + 14) ÷ 6 = 7 days
  • Uncertainty = (14 − 4) ÷ 6 ≈ 1.7 days

So you'd say "about 7 days, likely between 5 and 9". That's more honest than "6 days".

Team estimation: Wideband Delphi and planning poker

Several people estimate independently, then discuss the differences, then estimate again. Planning poker, common in Agile teams, is a version of this with cards.

The value isn't the number. It's the discussion: when one person says 2 days and another says 10, one of them knows something the other doesn't.

Communicating the estimate

An estimate is a forecast, not a promise. Present it with:

  • A range, not a single number.
  • Assumptions: "assumes the test environment is ready on day 1 and the payment sandbox is available".
  • The biggest risks to the estimate, and what happens if they occur.

Then re-estimate as you learn. A week into testing, you know much more than you did when planning.

Real-world example'Two weeks' became five

A test lead estimated two weeks to test a new invoicing module, based on the size of the feature.

Testing took five. The code worked fairly well. The time went elsewhere: the test environment wasn't ready for four days; the tax service sandbox was down twice; and each round of bug fixes needed a full redeploy that took half a day.

For the next release, the lead used three-point estimates and listed assumptions explicitly: "environment ready on day 1, sandbox available, automated deploys". The estimate was 3–5 weeks, with each assumption as a named risk. When the environment was late again, the delay was expected and visible, not a surprise.

Lesson: most estimation errors come from assumptions nobody wrote down. Name them, and the estimate becomes something the team can manage.

Do

"Testing should take 3 to 5 weeks. The main assumption is that the test environment is ready on day 1; each day of delay adds a day."

Don't

"Testing will take 3 weeks." Said once, it becomes a deadline, and the assumptions behind it are forgotten.

Check yourself

For a task, the optimistic estimate is 2 days, most likely 5 days and pessimistic 14 days. What's the three-point estimate, and what does the wide range tell you?

Show answer

E = (2 + 4 × 5 + 14) ÷ 6 = 36 ÷ 6 = 6 days. The wide gap between 2 and 14 shows high uncertainty. It's worth finding out why: unclear requirements, a risky dependency or new technology. Then tell the team about it rather than hide it in a single number.