StackLift
All terms

Glossary

Acceptance Criteria

Specific, testable conditions a feature or story must meet to be accepted as complete.

Acceptance criteria are the specific, testable conditions a feature or user story must meet to be accepted. They turn a general need into checks that a developer can build to and a tester can verify, as a pass or fail.

They are written before work starts, often as Given/When/Then statements or a short checklist, and cover error and empty cases as well as the happy path.

They differ from the definition of done, which applies to all work. A story is finished only when it meets both its own criteria and the shared definition of done.

Start in 60 seconds

Put this to work on your project.

Get an AI-assisted, architect-reviewed scope, cost range, and timeline. Free, no login.

The complete guide

Acceptance Criteria: the complete guide

Updated November 18, 2026 · Published by , Founder, StackLift AI. Written with AI assistance; how we make content.

Short answer

Acceptance criteria are the specific, testable conditions a feature or user story must meet to be accepted as complete. They turn a general need into checks a developer can build to and a tester can verify, so everyone agrees on what done means for that one item before work starts.

What are acceptance criteria?

In plain terms, acceptance criteria are the list of things that must be true before someone says, yes, this is what I asked for. They attach to a user story or a feature, and each line can be checked as passed or failed.

More precisely, they are a contract between the person who wants the work and the people who build it, scoped to a single item. The story says who wants what and why. The criteria say how anyone will know it works. A story without criteria leaves that question to be settled at the end, when time is short and opinions differ.

Good criteria describe observable behavior. They do not describe the code, the database or the framework. A tester who never saw the design discussion should be able to read them and run the checks.

Why do acceptance criteria matter for a software project?

Most disputes at the end of a build come from two people holding different pictures of the same feature. Written criteria put the picture on paper while changing it is still cheap. They also give developers a target, give testers a checklist, and give the client a fair basis for approval.

  • They reduce rework, because gaps surface in refinement rather than in review.
  • They make estimates more honest, because the team sees the real edge cases before sizing the story.
  • They keep acceptance objective: the work passes the agreed checks or it does not.
  • They separate defects from new requests. A missed criterion is a defect; a new need is a change.
  • They help new team members understand a feature without a long history lesson.

What formats can acceptance criteria take?

There is no single required format. Teams use whichever one the whole group will read and keep current. Three show up most often.

Common acceptance criteria formats
FormatHow it readsBest for
Given / When / ThenGiven a starting state, when the user acts, then this result happens.Behavior and flows; maps directly to automated tests
ChecklistA short list of true-or-false statements.Simple rules, limits, content and display needs
Example tableInputs and the expected output, one row per case.Calculations, validation rules, pricing and permissions

Given/When/Then comes from behavior-driven development, where the same lines can drive automated tests. A checklist is faster to write and fine for small items. Example tables are the clearest way to pin down rules that have many combinations.

What does a real example look like?

Take the story: as a buyer, I want to filter products by price, so that I can find what fits my budget. Here is a workable set of criteria in mixed formats.

  • Given a product list, when I set a minimum and maximum price, then only products in that range are shown.
  • Given a range with no matches, when I apply it, then I see a message and a way to clear the filter.
  • Given a filter is applied, when I reload the page, then the filter is still applied.
  • If the minimum is higher than the maximum, the form shows an error and does not search.
  • The filter works with the keyboard and with a screen reader.

Notice what is missing. There is no mention of the database query, the component library or the API shape. The criteria cover the happy path, the empty result, a persistence rule, an error and an accessibility need. That spread is what makes a short list useful.

How do you write good acceptance criteria?

The craft is mostly discipline. Follow the same steps for every story and the quality settles quickly.

  1. Write them before development starts, with the people who will build and test the story.
  2. Start from the user goal in the story, then ask what could go differently from the happy path.
  3. Make every line checkable: someone should be able to say pass or fail without debating it.
  4. Describe outcomes the user sees, not the internal implementation.
  5. Cover the error, empty, permission and slow-network cases, not only the happy path.
  6. Add the non-functional needs that matter for this story, such as accessibility, speed in plain terms, or audit logging.
  7. Keep the list short. If it runs long, the story is probably too large; split it.
  8. Review the criteria at refinement, agree on them, and update them if the story changes.

Acceptance criteria vs. definition of done

These two are often confused. Acceptance criteria belong to one story and say what it must do. The definition of done is a shared quality standard for all work, such as reviewed, tested, documented and deployed. A story is finished only when it meets both.

Acceptance criteria compared with related ideas
Acceptance criteriaDefinition of doneTest case
ScopeOne story or featureEvery item the team deliversOne specific check
AnswersWhat must this do?What quality bar must all work meet?What steps prove it?
Who sets itProduct owner with the teamThe team, often with stakeholder inputTesters and developers
ChangesPer storyRarely, by team agreementAs criteria change

A useful rule: if the line would be copied onto every story, it belongs in the definition of done. If it only makes sense for this feature, it is an acceptance criterion.

How do acceptance criteria help testing and sign-off?

Criteria give testers a ready-made list of checks. Many teams turn each Given/When/Then line directly into an automated test, so the specification and the test stay in step. Manual testers use the same lines as a starting script and then explore around the edges.

At sign-off, the product owner or client walks through the criteria and confirms each one. This keeps acceptance objective. If a user then asks for something that was not in the criteria, it is a new request, not a defect, and it is handled as a change. Writing that distinction down early avoids many disputes.

For larger releases, the same criteria feed a user acceptance testing plan, where business users confirm the work in realistic conditions before it goes live.

What mistakes should you avoid?

  • Vague words such as fast, user-friendly or robust with no way to check them. Replace them with an observable behavior.
  • Mixing in how to build it, which locks the team into one solution before they have explored others.
  • Writing them after the work is done, which only records what was built rather than what was needed.
  • Covering only the happy path, then discovering the empty state and the error state in production.
  • Adding new criteria late, which is scope creep in disguise. If the need is real, handle it as a change request with its cost and schedule effect visible.
  • Writing criteria nobody reads. If they live in a document the team never opens, they will not shape the work.

How do acceptance criteria differ across kinds of work?

The same idea applies everywhere, but what you check changes with the product.

What criteria focus on in different projects
Type of workWhat criteria usually cover
Web or mobile screenFields, validation messages, empty and loading states, keyboard and screen reader use
API or integrationInputs accepted, responses returned, error codes, retry behavior, access rules
Data or reportingWhich records are included, how totals are calculated, refresh timing, who can see what
AI featureWhat the assistant may and may not answer, handling of unknown questions, source citation, handoff to a person
MigrationWhich records move, how mismatches are reported, how rollback works

For AI features, plain pass-or-fail wording still works. Instead of saying the assistant is accurate, write that it answers from the approved documents, shows its source, and hands off to a person when it cannot find an answer. A small test set then checks each line.

How do you know your acceptance criteria are working?

You cannot measure criteria directly, but their effects show up in how a team works. Watch for these signals over a few sprints.

  • Fewer stories come back from review with the comment that this is not what I meant.
  • Testers start work from the criteria and rarely need to ask the developer what a story should do.
  • Disagreements at sign-off drop, and those that remain are about new requests rather than missed behavior.
  • Stories are split earlier, because a long criteria list is treated as a warning sign.
  • Defects found after release trace back to cases nobody listed, and the team adds those patterns to its checklist.

A short retrospective question helps: which bug or dispute this sprint would a better criterion have prevented? Collect the answers and the writing gets sharper without any extra process.

What tools do teams use for acceptance criteria?

The tool matters less than the habit. Most teams keep criteria inside the story in their tracker, such as Jira, Linear or Azure DevOps, so the text travels with the work. Teams that automate checks often use Cucumber, SpecFlow, Behave or Playwright and Cypress tests that read Given/When/Then lines.

If you are starting out, a plain checklist in the story description is enough. Add automation once the same criteria are being checked by hand every release and the repeated effort starts to hurt.

What do acceptance criteria mean for cost and timeline?

Criteria are the most practical way to size a story. Each extra edge case, permission rule or integration check adds real work, and listing them early makes the estimate more accurate. Vague stories tend to hide the effort until it shows up as delay.

That is why StackLift prices by deliverable complexity rather than hours, with scope agreed before engineering begins. The free estimate scopes a project feature by feature, and a 3-day AI Scope Sprint produces detailed scope and user stories you can attach criteria to. See software estimation playbook for how scope turns into a plan.

How does StackLift handle acceptance?

Every engagement starts with structured discovery and architect-reviewed scope, so the work is defined before it is built. Estimate, proposal, project and client portal share one baseline, which means a change to what was agreed is explicit rather than silent.

Clients see live project state, approvals, invoices and releases in the client portal, so accepting work against agreed criteria is visible to everyone. AI agents and senior engineers do the build work, and engineers own the decisions about what meets the bar.

Questions people ask

Who writes acceptance criteria?

Typically the product owner or business analyst drafts them, and the developers and testers review them. Writing them together catches gaps early and gives everyone the same picture.

How many acceptance criteria should a story have?

Only as many as it takes to define done, often a handful. A very long list signals that the story should be split into smaller ones.

Are acceptance criteria the same as test cases?

No, but they are closely linked. Criteria state what must be true. Test cases are the steps that check it, and a criterion often becomes one or more tests.

Who accepts the work?

The product owner or the client representative decides whether the criteria are met. Developers and testers verify first; the owner makes the final call.

Can acceptance criteria change during a project?

They can, with agreement. Small clarifications are normal. A change that adds work should go through a change request so the cost and schedule effect is clear.

Do I need acceptance criteria if I am not using Agile?

Yes. Any project benefits from testable conditions per feature. In a fixed-scope contract they are often the basis for milestone sign-off, which makes them even more useful.

Should acceptance criteria include design details?

Include only the design points that change behavior or meaning, such as which fields are required. Pixel-level styling belongs in the design file or a style guide, not in each story.

What is the difference between acceptance criteria and requirements?

Requirements describe a need at a broad level. Acceptance criteria are the precise conditions that show a specific story meets part of that need. One requirement often produces several stories with their own criteria.

Sources

Sources and further reading

The standards, rules, research and official documentation this page draws on.