INSIGHT · ALASTAIR MAJURY

Acceptance criteria workshops: test the boundary, not just the happy path

Use contrasting examples, unresolved questions and evidence owners to make an acceptance-criteria workshop useful to business and delivery teams.

By Alastair Majury · Published · 5 minute read

A team can agree that a requirement is clear while imagining several different behaviours. Everyone supports the happy path: a valid request arrives, the system processes it and the user sees the expected result. The disagreement emerges later, when a request is incomplete, arrives twice or changes while somebody is reviewing it.

An acceptance-criteria workshop is a good place to expose those differences. Its purpose is to establish shared expectations about observable outcomes. The approach below is a practical facilitation method for analysts and delivery teams. It does not assume a particular delivery framework or replace the detailed testing that a system may need.

Begin with one user outcome

Choose a small, coherent piece of behaviour. Ask who needs it, what they are trying to achieve and why that matters. Read the proposed requirement aloud and invite an operational colleague to explain the intended outcome in their own words. Differences in language can reveal differences in understanding before anyone writes a test.

Keep the discussion anchored to that outcome. If a requirement contains several independent goals, split the conversation. Trying to settle every rule for an entire process in one workshop can create apparent agreement while leaving important dependencies unresolved. A narrower session makes it easier to recognise which questions belong elsewhere.

Put two contrasting examples beside each other

Start with an ordinary example, then change one relevant condition. What happens if the request arrives after the agreed business boundary? What if a required value is absent? What if the user submits the same request again? Choose examples that reflect the actual process rather than inventing an exhaustive catalogue of technical edge cases.

For a fictional internal request service, a team might agree that an approved request creates one work item. A second example asks what should happen when the same request is submitted twice. The answer might depend on how the business identifies duplicates and whether a user can legitimately repeat an earlier request. The workshop has now uncovered a decision, rather than merely added another sentence.

Ask what an observer would see

Replace vague statements such as handled correctly with a result that a user or authorised reviewer could recognise. Describe the status, message, work item or recorded outcome that demonstrates the expected behaviour. Where timing matters, define the relevant event and how the team will measure the interval.

Be careful with apparently precise numbers. A target entered during a workshop is not automatically an agreed service commitment. Confirm the source of the target and who can approve it. If the team lacks that information, preserve the question rather than filling the gap with a convenient assumption. Precision should reflect a decision, not conceal its absence.

Separate business rules from design suggestions

People often describe a preferred screen or button while trying to explain a business need. Capture the suggestion, then ask which outcome or constraint it supports. A rule about who can authorise a request may remain valid across several interface designs. A particular layout may be only one possible way to help that person act.

This distinction allows designers and developers to contribute without losing the business boundary. Some implementation constraints will be essential, especially where the existing environment limits the options. Record those constraints and their evidence explicitly. Avoid treating every familiar feature of the current process as a mandatory requirement for the future one.

Give unanswered questions an owner

A workshop does not need to resolve everything to be useful. It does need to make uncertainty visible. Record the question, the decision it blocks, the person responsible for obtaining an answer and when the team needs it. Identify any dependent stories or process steps that could change when the answer arrives.

Read the open questions back before closing. Agreement to investigate is different from agreement about behaviour. Keep unresolved conditions out of any claim that the requirement is ready for acceptance. When an answer arrives, update the examples and tell the affected people which expectation changed and why.

Agree how acceptance will be demonstrated

For each criterion, identify the evidence that would support acceptance and who should review it. A demonstration may be enough for some behaviour; other outcomes may need a report, a reconciliation or a recorded business review. Choose evidence proportionate to the consequence of getting the behaviour wrong.

The GOV.UK Service Manual explains user stories through the actor, need and goal, and describes acceptance criteria as outcomes that help confirm a need has been met. That guidance is written for government services. The workshop sequence here is a suggested professional practice, not a requirement imposed on every organisation.

Finish by asking participants to explain both the ordinary example and the contrasting one. If they describe the same outcomes, the workshop has produced something useful: a shared boundary that delivery and business reviewers can examine together.

Related reading: Before requirements.

Source and scope

GOV.UK Service Manual: Writing user stories. Published 16 February 2016; last updated 23 May 2016. Checked 20 November 2026.

Back to all insights