Acceptance Criteria: 2 Formats and 5 Checks Every Business Analyst Should Know
"The discount code should work properly." Is that an acceptance criterion? Not one anybody can test. In this video I take one vague user story from a made-up online shop and rewrite its acceptance criteria from bad to good: what acceptance criteria really are, the 2 formats (rule-based and Given/When/Then), 5 checks for testable wording, the finished list with scenarios for the tricky cases, and the 5 mistakes to avoid.
Step by step
Acceptance criteria are the finish line
- Acceptance criteria are the conditions one user story must meet to be accepted. The definition of done is different: it covers every story. Our example story: as a shopper, I want to enter a discount code, so that I pay less. Its only criterion, "work properly and be easy to use", can't be tested.
Format 1: rule-based criteria
- Format 1 is rule-based: a checklist where each line is one rule that is true or false. For example: a valid code takes 10% off the items in the cart, and only 1 code can be used per order. Rules suit limits, amounts and conditions, but they don't say what the shopper sees when a rule is broken.
Format 2: Given, When, Then
- Format 2 is Given/When/Then, from behaviour-driven development. You write one concrete example, called a scenario. Given a cart with items worth $80 and a valid 10% code, When the shopper applies the code, Then the cart shows a discount of $8 and a new total of $72.
A common mistake: the click script
- Don't write a scenario as a click script: open the shop, add a mug, click Apply, see a pop-up. That has lots of steps and no rule, so the 10% is never written down. Keep to 1 situation, 1 action and 1 result, and leave the clicks to the tester.
Which format to pick
- As a rule of thumb, use rules for simple limits and conditions, like 1 code per order or items of $50 or more. Use Given/When/Then when the behaviour depends on the situation, like an expired code. You can mix them: a few rules plus a scenario for each tricky case.
Explain like I'm 5: clean your room
- Think of a parent who says "clean your room". Nobody said what clean means, so they argue. Now try: toys in the box, clothes in the basket, bed made. Each one is true or false. Acceptance criteria do the same job: you agree what acceptable looks like before the work starts.
Acceptance vs evaluation criteria (BABOK)
- The BABOK Guide from the IIBA has a technique called acceptance and evaluation criteria. Acceptance criteria are typically for 1 solution and are usually pass or fail. Evaluation criteria rank several options, like 3 payment providers. The guide asks for a testable form.
When to write them, and who with
- Write them before anyone builds, and not alone. Book 30 minutes before the story is planned with 3 points of view: the business (here the sales manager), a developer and a tester. Bring a first draft and let them poke holes in it. Criteria can change later if everyone agrees.
Check 1: true or false?
- Check 1: can it be true or false? Ask whether 2 people could disagree about whether it passed. "The discount code works properly" fails this check. "A valid code takes 10% off the items in the cart" passes. Ask a tester: could you fail this?
Check 2: numbers, not adjectives
- Check 2: numbers, not adjectives. Swap words like fast, easy and properly for something you can count or see. "Appears fast" becomes "appears within 2 seconds". Get the numbers from the people who own them: here the sales manager picks 10% and finance sets the $50 minimum.
Check 3: one thing each
- Check 3: one thing per criterion. "The code takes 10% off, shows the discount as its own line, and updates the total within 2 seconds" is 3 checks in 1 line. Split it into 3 lines so each can pass or fail alone. When you see the word "and", ask if it should be 2 criteria.
Check 4: the outcome, not the design
- Check 4: say the outcome, not the design. "The shopper clicks the green Apply button and a pop-up shows the discount" is a screen design. Write: the shopper can apply a code from the cart page and sees the discount before paying. Only fix a detail that really is a rule, like error message wording.
Check 5: the unhappy paths
- Check 5: cover the unhappy paths. Ask what can go wrong: the code has expired, it was typed wrong, the shopper tries a second code, or the items come to less than $50. For each case, write what the shopper sees and what happens to the total, so the developer doesn't have to guess.
Worked example: the bad to good rewrite
- The vague line became 6 rules: a valid code takes 10% off the items; the discount shows as its own line before payment; it does not apply to shipping; only 1 code per order; the items must come to $50 or more before the discount; the new total appears within 2 seconds.
Scenarios for the tricky cases
- Each tricky case gets a scenario. Expired code: Given items worth $80 and a code that expired yesterday, When the shopper applies it, Then the cart shows "This code has expired" and the total stays at $80. For a second code, the cart keeps the first and shows "Only one code per order".
Confirm it, and where it goes
- Take the list back to the same 3 people. You are done writing when the sales manager says it's what they meant, the developer can build it and the tester can check it. Keep the criteria with the user story, for example in Jira or Confluence. The same list is used again in user acceptance testing.
5 common mistakes
- 5 common mistakes: fog words like fast, easy and properly; writing the design instead of the outcome; covering only the happy path; writing them alone after the build has started; and writing too many. A very long list is a sign the story is too big, so think about splitting it.
Recap
- In short: acceptance criteria are the conditions one story must meet. Use rules for limits and conditions, and Given/When/Then for behaviour and tricky cases. Run the 5 checks: true or false, numbers not adjectives, one thing each, outcome not design, unhappy paths. Write them before the build.
Common mistakes
- Fog words like fast, easy and properly. Nobody can test them. Swap each one for something you can count or see, such as "the new total appears within 2 seconds".
- Describing the screen instead of the outcome. A criterion about a green button breaks as soon as the layout changes. Write what must be true, such as "the shopper sees the discount before paying", and let the designer choose how.
- Only the happy path. Add a scenario for each thing that can go wrong, like an expired code or a second code. Say what the shopper sees and what happens to the total.
- A Given/When/Then that is a click script. A list of clicks hides the rule, so the 10% never gets written down. Keep 1 situation, 1 action and 1 result, with real numbers like $80 and $72.
When a business analyst uses this
Before a story is planned, a business analyst brings draft criteria to a 30-minute chat with the business (for example the sales manager), a developer and a tester, then saves the agreed rules and scenarios on the story. The tester runs each line, and the business checks the same list in user acceptance testing.
Questions people ask
Is acceptance criteria the same as the definition of done?
No. Acceptance criteria are the conditions one user story must meet to be accepted. The definition of done is the team's list that covers every story.
What is the Given/When/Then format for acceptance criteria?
It is one concrete example in 3 parts: Given is the starting situation, When is the one thing that happens, and Then is the result you can see. For example: Given items worth $80 and a valid 10% code, When the shopper applies it, Then the new total is $72.
Can acceptance criteria change after they are written?
Yes, as long as everyone agrees to the change: the business, the developer and the tester. Changing them quietly after the build has started turns an agreement into a surprise.