A vendor saying the build is done is not the same as the build working for your business. Six steps help non-technical founders test against real tasks, record defects and sign off with evidence.
The message arrives on a Thursday afternoon: "Build complete. Please confirm acceptance so we can invoice the final milestone." You open the app, click through the main screens for twenty minutes, find nothing alarming and reply with a thumbs-up.
That reply is a contract moment, and most founders treat it as a courtesy. A proper round of user acceptance testing would have asked different questions. Can a customer actually pay? Can staff reverse a mistake? Does the report match the spreadsheet it replaces? Those questions are cheap on Thursday and expensive in March.
Acceptance testing is not a technical skill. It is a structured way of using the product the way your business will, writing down what breaks, and deciding calmly whether the work meets what was agreed. Developers cannot do it for you, because they know too well how the software is meant to be used.
Six steps follow, from turning the scope into pass-or-fail statements to signing off in writing. Before the first, it helps to separate acceptance testing from the other kinds of testing a vendor will mention.
What is user acceptance testing, and who is it for?
User acceptance testing, or UAT, is the final check in which the people who will use or pay for software test it against real tasks and agreed criteria before accepting delivery. It differs from developer testing, which hunts technical faults. UAT asks a simpler question: does this build do the job the business bought it for?
Vendors often use three testing terms loosely, so it helps to see who owns each one.
| Test type | Who runs it | Question it answers | When it happens |
|---|---|---|---|
| Developer testing | The developers | Does each piece of code do what it should? | During the build |
| QA testing | A tester, usually inside the vendor team | Does the whole product work and hold up under use? | Before delivery |
| User acceptance testing | You and the people who will use the product | Does it do our job, as agreed? | At delivery, before sign-off |
The first two are the vendor's responsibility. The third is yours, with the vendor's support, which is why it is the one most often skipped or rushed.
An illustrative buyer
Picture a composite founder, invented for this article, who runs a small training academy and has commissioned a course-booking portal. The vendor reports the build complete. The founder logs in as an admin, adds a course, books one seat and approves.
Weeks after launch, the first customer asks for a refund and nobody can find where to issue one. The steps below show how a planned round of acceptance testing would have surfaced that gap while it was still a small fix.
1. What does 'accepted' mean for each feature?
Acceptance criteria are short pass-or-fail statements that define when a feature counts as done, such as 'a customer can cancel a booking and receive a confirmation email'. Written before testing begins, they stop acceptance from turning into opinion. If a criterion cannot be checked by someone following simple steps, it is too vague to use.
What to do
- Take the agreed scope or feature list as your starting point.
- Write one to three statements per feature in the form 'a user can do something and this result follows'.
- Add the unhappy path: what happens when an entry is wrong, a payment fails or the user changes their mind.
- Ask the vendor to confirm each statement in writing before the build is declared complete.
What good looks like
Every feature has criteria that a non-developer can check, the whole list fits on a few pages, and both sides have agreed it. Criteria that exist only in your head are preferences, not terms.
Watch out for
Criteria written after delivery give the vendor a fair argument that the scope was different. Phrases such as 'works well' or 'is fast' need a measure you can observe. If the scope itself is loose, tighten it first, and comparing vendor scope line by line shows where the gaps are.
Delivery day coming with no test plan?
If acceptance still means a quick click-through, the first fix is a written definition of done.
Define your done2. Does the test environment look like real life?
A test environment, often called staging, is a private copy of the product running separately from the live one. For acceptance testing it needs realistic data and the same settings as production, including payments in test mode and email that reaches a test inbox. Testing on an empty system proves little, because real work involves messy, varied data.
What to do
- Ask for a staging link that is separate from the live product.
- Load it with realistic sample data, such as real product names, awkward customer records and long text, anonymised where personal data is involved.
- Create a test login for every role: admin, staff and customer.
- Connect payments, email and SMS in their test modes.
- Confirm that nothing done in staging can reach a real customer.
What good looks like
You can log in as every role, the data resembles your business, and a mistake costs nothing.
Watch out for
Testing on the live system, where a test refund can become a real one, is the classic error. A staging copy that differs from production in settings is the other, because the bug then appears only after launch. Ask the vendor in writing how the two environments differ.
3. Who should test, and what should they actually do?
Acceptance testers should be the people who will use the product daily, plus one decision-maker who can accept or reject. Give them scripts built from real jobs, not a tour of features. A good script reads like an ordinary working day: take an order, fix a mistake, produce the month-end report.
What to do
- Choose two to five testers, including one sceptical colleague who will push on weak spots.
- Write five to ten scripts from real tasks, in plain language, each with an expected result.
- Include hand-offs between roles, such as staff creating a request and a manager approving it.
- Include a script of deliberate mistakes: a wrong entry, an undo, a cancellation, a refund.
- Set a testing window and protect the testers' time during it.
What good looks like
Testers finish tasks without asking the vendor how, and they report friction as well as outright failures.
Watch out for
The person who commissioned the project often tests alone, and knows exactly where everything is, which hides the problems a new user would meet. Testing only the happy path is just as common. If a developer sits beside a tester explaining each step, treat the need for that help as a finding.
4. How do you record a defect so it actually gets fixed?
A defect report should let a developer reproduce the problem without a phone call. It needs the steps taken, what was expected, what actually happened, the user role and device, and a screenshot or screen recording. Vague notes such as 'checkout is broken' cost days of back-and-forth, while a clear report is often fixed in one pass.
What to do
Use one shared sheet or tracker, with one defect per row and these columns:
- A short title naming the screen and the problem
- Numbered steps to reproduce it
- Expected result and actual result
- Evidence: screenshot or screen recording
- Where it happened: role, browser or device, date and time
What good looks like
Each entry stands alone, and someone who was not in the room could reproduce it.
Watch out for
Feedback scattered across email, chat and calls means nothing is counted and some of it is lost. Reporting several problems in one message is the other trap. Give each defect its own entry so each can be closed on its own.
5. How do you decide which defects block acceptance?
Triage sorts defects by their effect on the business, not by how annoying they are. A blocker stops a core task or puts money, data or customer trust at risk. A major defect has a workaround but hurts daily work. A minor defect is cosmetic or rare. Agreeing these classes in advance keeps the conversation with the vendor calm.
What to do
Give every defect one class, then one decision:
- Blocker: a core task fails, or money, data or customer trust is at risk.
- Major: the task can be completed through a workaround, but daily work suffers.
- Minor: cosmetic, rare or low impact.
- Not a defect: a new idea, which goes on the roadmap list.
The decision is one of three: fix before acceptance, fix within an agreed period after acceptance, or move to the roadmap.
What good looks like
Every defect has a class and a decision, and the vendor has agreed with the classes.
Watch out for
New ideas often arrive during testing and get logged as defects, which stops the build ever being finished. Keep them on a separate list, and if that list keeps growing, see how to keep feature creep out of the build. Treating every defect as equally serious is the opposite mistake, and it turns a sound delivery into a fight.
6. What does a proper sign-off record look like?
A sign-off record is a short written statement that names the build version, lists the tests run, records open defects with agreed fix dates, and states whether you accept, accept with conditions or reject. It matters because acceptance usually triggers the final payment, the start of the support period and the handover of code and accounts.
What to do
- Retest every fixed defect and rerun the key scripts, because a fix can break a neighbouring feature.
- Record the build version and the date of testing.
- List every open defect with its class and an agreed fix date.
- State the decision and any conditions.
- Have both sides sign, and keep the record where your successor could find it.
- Check your contract for what acceptance triggers, including payment, warranty and support.
What good looks like
A one-page record that a colleague joining next year could understand without asking anyone.
Watch out for
Informal acceptance by chat, such as 'looks good to me', can later be read as full acceptance. Approving to hit a launch date without recording the open defects is the other common slip. This article is not legal advice, so ask a lawyer what your contract says acceptance triggers.
Where the effort goes, and what skipping a step costs
The table compares the six steps on the effort each takes and the risk you carry if it is skipped.
| Step | Effort | Risk if skipped | Usually owned by |
|---|---|---|---|
| 1. Acceptance criteria | Low | High | You and the vendor, together |
| 2. Test environment and data | Medium | High | Vendor sets up, you check |
| 3. Testers and scripts | High | High | You and your team |
| 4. Defect recording | Low | Medium | The testers |
| 5. Triage | Low | High | You and the vendor, together |
| 6. Retest and sign-off | Medium | High | You decide, the vendor fixes |
The cheapest steps, writing criteria and triaging defects, guard against the costliest disputes, while tester time is the real expense, so book it before delivery day.
Bring your build and your scope for a second opinion
Share the agreed scope and the delivered build, and we will help you shape a test plan and a fair sign-off.
Review your acceptance planAccept, accept with conditions or reject: how to decide
Accept when no blockers remain and the key scripts pass. Accept with conditions when open defects are minor or have workarounds and the vendor commits to dated fixes in writing. Reject when a core task fails, data is at risk or the build does not match the agreed scope. Record whichever decision you make.
A build is accepted when it passes the job you agreed, not when everyone is tired of testing.
When a light version is enough
A small change with low risk, one user role and no payments or personal data can be checked in an hour against a short list of criteria. Keep the written record anyway, because it costs little and settles later questions.
When full acceptance testing is essential
Run the full six steps when the product handles payments or personal data, when several roles hand work to each other, when it replaces a manual process people depend on, or when data is being moved from an old system.
When outside help earns its place, and when it does not
An independent tester helps when nobody on your side can write scripts, when the system is regulated or high-value, or when trust with the vendor has worn thin. It is unnecessary when the vendor supplies a clear test plan and your own staff have the time and the knowledge of the work.
Outside testers can find faults, but only you can decide whether the product fits your business.
How KSoft Technologies Helps With Acceptance and Sign-Off
We build SaaS and MVP products, and our projects begin with a written scope, which gives acceptance criteria something firm to stand on. Founders who start from that scope usually find the six steps above much easier, because most of the definition of done already exists on paper.
We can also help teams who have inherited a build from another vendor. That usually means turning the scope into criteria, setting up a realistic test environment, running a first round of testing with your people and preparing a sign-off record before any final payment or handover.
If you would like to talk through how to test a delivered build, bring the scope and a staging link and we can start there. More from our team is on the KSoft Technologies YouTube channel.
Treat the sign-off as a decision you can defend
Acceptance testing looks like a technical chore and works like a business control. It converts a vendor's claim of completion into evidence you can weigh, with named people, real tasks and a written result.
The pattern across the six steps is the same. Define done before the build ends, test on something that looks like real life, use the people who do the work, write defects so they can be fixed, sort them by business impact and keep a record. Skip any one and the sign-off becomes a guess.
A sensible next step costs one afternoon. Take your current or next project, list the three tasks that cannot fail, and write each as a numbered script with an expected result. Share them with the vendor now, while there is still time to adjust the plan.
When delivery day arrives, you will be accepting evidence instead of goodwill, and that is what keeps a hard conversation with a vendor short and civil.
Sign off with evidence, not hope
A written test plan and a dated defect list make the final payment a decision you can defend. Talk to us about setting both up before delivery.
Plan your sign-offFrequently Asked Questions
What is the difference between UAT and beta testing?
UAT is a planned check by named people against agreed criteria, done before you accept a delivery. Beta testing releases a near-final product to a wider group of real users to find problems in everyday use. UAT decides whether to accept the work. Beta decides what to fix before a broader launch.
Why do vendors and clients disagree about whether software is finished?
Usually because 'finished' was never defined in testable terms. The vendor reads the scope as a feature list, while the buyer reads it as an outcome. Without written acceptance criteria, each side relies on memory and goodwill. A written definition of done, agreed before the build ends, removes most of the argument.
Why does skipping acceptance testing cost more later?
Defects found after launch affect real customers and real data, and the fix competes with new work. Disputes also get harder once the final payment is made and the vendor has moved on. Finding the same issue during acceptance testing costs a short fix cycle, with no customer impact and clearer standing.
Who should take part in user acceptance testing?
Include two to five people who will use the product daily, one person with authority to accept or reject, and a vendor contact who answers defect questions. Add one sceptic who tends to find problems. Avoid testing by the project sponsor alone, because familiarity hides friction that new users hit straight away.
Should I test on the live system or a staging copy?
Use a staging copy, a private replica of the product, for testing. Live systems risk real emails to customers, real payments and polluted data. The staging copy must match production settings and hold realistic sample data. Do a brief final check on the live system after release, limited to harmless actions.
Can a non-technical person run acceptance testing alone?
Yes, for most business software. Acceptance testing needs knowledge of the work, not of code. Write scripts from real tasks, log defects with steps and screenshots, and ask the vendor to explain technical causes. Bring in outside help only when the product handles regulated data, complex integrations or payments you cannot judge yourself.
When is the right time to start preparing for UAT?
Start preparing when the scope is agreed, long before delivery. Write acceptance criteria then, book testers' time in the calendar, and ask for a staging link early. Waiting until the vendor says 'done' squeezes the test into days, invites rushed approval and leaves no time for a second round after fixes.
How do I run a first acceptance test this week?
Pick the three tasks your business cannot do without, write each as a short numbered script with an expected result, and run them on a staging copy with one colleague who did not commission the work. Log every problem in a shared sheet with steps and screenshots. That small run reveals most gaps.
How long does user acceptance testing take?
It depends on product size, risk and how many roles use it. Plan a testing window, a fix period and a retest round, not a single pass. Small changes may need hours, while a new multi-role system needs days of focused testing. Ask the vendor to agree the window in the project plan.
What happens if I accept a build and find bugs afterwards?
It depends on your contract. Many agreements include a support or warranty period for defects found after acceptance, but terms vary, and a vague approval can weaken your position. Log problems immediately with evidence, refer to the sign-off record and contract clauses, and seek legal advice for disputes. This is general information, not legal advice.
How much does acceptance testing cost?
The main cost is your team's time, plus any vendor hours to fix defects, which the contract may include or bill separately. No standard figure exists. Reduce cost by writing criteria early and logging clear defects, since vague reports multiply fix cycles. Ask in advance how many retest rounds the price covers.
Should I hire an outside tester or do UAT in-house?
Do it in-house when your staff know the work and have time. Hire an independent tester when the system is complex, regulated or high-value, when your team cannot spare time, or when trust with the vendor is low. An outside tester complements your own acceptance, because only you can decide fitness for your business.
