The new system looks ready in the demonstration. A request moves to an approver, stock appears on a dashboard and a report downloads. Then your operations manager asks: what happens when the approver is away, there is only part of the stock, or the connection drops after someone clicks submit?

Those questions belong in the buying process. Software acceptance testing gives the people responsible for the business a way to check the agreed work before accepting a release. User acceptance testing, usually shortened to UAT, asks whether intended users can complete their real tasks with acceptable results. It needs to sit alongside the delivery team’s technical, security and other testing.

For a Kenyan organisation commissioning custom software, replacing a spreadsheet workflow or preparing a web or mobile application for launch, the useful outcome is a documented decision: which requirements have passed, what remains unresolved and who is authorised to accept the remaining risk.

Approaching a software demonstration or handover?

Agree what your team needs to prove.

Describe one important workflow, the roles involved and the failure you most need to prevent. Autocode Studios can help turn that into a practical delivery and acceptance scope.

Discuss the acceptance scope

Start with the business journey you are accepting

For example, imagine a distributor introducing an internal purchasing system. A branch employee requests supplies, a manager approves the request and a storekeeper records what was issued.

A feature list might say “requests, approvals and inventory”. The business journey is more specific: a branch requests twelve units, eight are available, the manager approves a partial issue and the remaining four stay visibly outstanding. The requester, manager and storekeeper need to see the same state without losing the original request.

That single journey exposes several acceptance questions. Who may change the quantity? Can someone approve their own request? What does the requester see after a partial issue? Does the report count requested units or issued units? Use your team’s actual rules to answer these before testing starts. If the first release is still being defined, the guide to agreeing an MVP scope helps establish the boundary.

Write the expected result before the demonstration

An acceptance criterion describes an observable outcome. For the distributor, a useful criterion might be: after an authorised storekeeper issues eight units against an approved request for twelve, the request shows eight issued and four outstanding, stock decreases by eight and the issue records the responsible user and time.

The accompanying test describes the starting records, the person’s role, the steps and the evidence to inspect. Keep the criterion focused on the business result. Let the delivery team explain the implementation and identify technical checks needed to support it.

Agree who can approve a criterion and how changes will be handled. Discovering a missing agreed rule is different from requesting a new capability. Record both, then use the project’s agreed change process to resolve scope and delivery implications.

Follow the same request through the awkward cases

The ordinary journey is only the beginning. Keep following the purchasing request when something changes. The most useful scenarios come from the people who already handle exceptions at the branch, in stores and in finance.

  • Approval changes: the manager is absent, the request is rejected or a permitted delegate acts. Verify the agreed authority and the resulting notification.
  • Permission boundaries: a requester attempts self-approval or opens another branch’s restricted record. Check that the action is denied and sensitive information remains protected.
  • Partial fulfilment: eight of twelve units are issued, then the remaining four arrive later. Confirm the request history, balance and report agree after both issues.
  • Interrupted submission: the connection drops after an issue is submitted. Verify the user can establish whether it succeeded and cannot accidentally record the same issue twice.
  • Correction: an authorised person reverses an incorrect issue. Confirm the stock and reporting adjustment follows the agreed rule and retains the relevant history.

For each case, state what should happen and what must never happen. The role and permission guide helps clarify who may act; the audit-trail guide explains the evidence an important action may need. Where another system is involved, agree how both sides will be reconciled using the integration planning guide.

Give real users a controlled place to test

Choose representatives from the roles in the journey, including someone who handles exceptions. A manager’s demonstration cannot establish whether a storekeeper can complete the work on the device used at the counter. Give each tester an appropriate account and let them carry out the task.

Use a separate test environment with representative, authorised test data. Match the production conditions that matter: application version, role settings, device sizes, browsers and supported integrations. Document differences, including simulated payments, messages or third-party responses. A simulation can test part of a journey, but it cannot establish that the production connection has been configured correctly.

Keep real orders, payments and outbound messages out of rehearsal unless a separately authorised production verification is planned. Use synthetic or appropriately protected data, and restrict access to test evidence. There is usually no need to circulate customer records in screenshots or enquiry messages.

Record evidence that leads to a decision

In the distributor example, the eight-unit issue might look right on screen while the monthly report counts twelve. “The stock page works” would miss the disagreement. A useful test record connects the requirement to the screen, resulting records and report totals that prove the complete outcome.

Keep one register with the scenario reference, related requirement, tested release, role, starting data, expected result, observed result, evidence reference, tester and date. Record whether the test passed, failed or could not be completed. An untested scenario should stay visible rather than being counted as a pass.

A failure report should explain the business impact and the steps needed to reproduce it. Agree severity together: a flaw that permits an unauthorised issue is materially different from a label that needs clearer wording. After a repair, retest the failed case and relevant neighbouring journeys; changing partial fulfilment may also affect reversals and reports.

Check the conditions for daily operation

Once the request journey behaves correctly, ask whether people can use and support it in the working environment. Test essential tasks on the agreed mobile and desktop devices, with keyboard navigation, understandable errors and readable content. Agree measurable response expectations for important tasks under a representative workload instead of accepting “fast” as a requirement.

Ask the delivery team for evidence of the security, performance and recovery checks appropriate to the system. UAT by business users does not establish that every technical risk has been addressed. Operational readiness may also require a demonstrated restore, support access, monitoring, training and a fallback procedure if launch is interrupted. The software handover guide helps turn those responsibilities into deliverables.

Make acceptance and launch explicit decisions

Before the test period, agree the conditions for accepting the release. They should cover critical journeys, unresolved defects, completed technical checks, business approval and operational readiness. Avoid a universal pass percentage: many successful cosmetic checks cannot compensate for a failed permission boundary or incorrect stock balance.

At review, name the release being considered and the evidence examined. A defensible decision may accept it, accept a clearly limited scope with documented exceptions, or require further work. For every accepted exception, record its impact, temporary control, accountable owner and agreed resolution date. Only someone with the right business authority should accept that risk.

Acceptance, deployment and payment milestones can have different meanings under the project agreement. Make their relationship explicit with the relevant project and commercial owners. Do not assume that signing a demonstration attendance sheet means accepting every deliverable or that a UAT pass alone authorises a production launch.

The distributor can now make a practical choice. If partial issues and permissions pass but one non-essential report still needs work, the owners can assess a limited release against a documented workaround. If stock totals cannot be reconciled, the evidence explains why further correction is required. Either way, the decision follows the business consequence.

Common questions before software sign-off

Who should carry out user acceptance testing?

People who understand the real business tasks should test representative journeys, supported by the delivery team. Name a business owner with authority to decide whether the evidence meets the agreed acceptance conditions.

Is a successful software demonstration enough for acceptance?

A demonstration can explain the system, but acceptance needs agreed criteria and recorded evidence. Intended users should verify important journeys, exceptions, permissions and resulting records in the agreed test conditions.

Must every defect be fixed before launch?

The decision depends on the agreed acceptance conditions and business risk. A limited exception needs a documented impact, temporary control, accountable owner and resolution plan. Critical failures must be assessed against the agreed launch blockers.

Does UAT replace security and performance testing?

No. Business acceptance, security, performance and operational readiness address different questions. Request the appropriate technical evidence alongside the results from business users.

Make the next demonstration decision useful

Bring one workflow and its hardest exception.

Tell Autocode Studios what your team needs to complete, which roles take part and what would make the release unsafe to accept. We can discuss a suitable scope for custom software development, workflow automation or software improvement and support.

Plan the next software review Use the contact page