Mon – Sat: 9:00 AM – 7:00 PM IST
Implementation

What is UAT (user acceptance testing)?

The phase where the business confirms the system does its job — and the one most often reduced to a signature.

In short

User acceptance testing is where the people who will use the system run their real processes against it, with real data, and confirm it works. It is not a bug hunt — defects should have been found in system testing. It answers a different question: can this business run on this configuration?

Why It Matters

Why this matters in practice

UAT is routinely compressed into the last two weeks and reduced to a sign-off. When that happens the business discovers during the first live month what it should have discovered in UAT, at far higher cost and with customers watching. The value is not in the testing hours; it is in the business committing that the configuration matches how it actually works.

Run by business users, not by the project team
Uses real data and real end-to-end processes
Confirms fit, not whether the software functions
Exit criteria are defined before it starts
How It Works

How it works in practice

Write scenarios, not clicks

"Process a credit customer's rush order with a partial despatch" — a real situation, not a sequence of button presses.

Use migrated data

Testing against sample records proves nothing about whether your actual data works in the configuration.

Cover the awkward cases

Returns, credit notes, part-deliveries, cancellations. Happy paths rarely fail; exceptions do.

Log and triage

Defects separated from change requests. Conflating them is how scope grows during UAT.

Exit on criteria

Defined before UAT begins — for example, no open critical defects and every core scenario passed by a business owner.

Worked example

A worked example: why UAT passed and go-live failed

A UAT cycle completes with 94 percent of scripts passed and sign-off given. Two weeks after go-live the business is struggling, and the obvious question is how testing missed it.

It did not miss it; it was not looking. The scripts covered the standard path of each process — a clean order, a standard receipt, a normal invoice. What broke in production were the variations: part deliveries, credit notes against partly paid invoices, orders with mixed tax treatment, the month-end sequence under time pressure. None of those were scripted, because scripts were written from process documentation rather than from what actually happens.

Two practices prevent this. Write scripts from real transactions pulled out of the legacy system, including the awkward ones, rather than from the documented happy path. And test the periodic processes — month-end, payroll, statutory filing — at least once in full, because they are the highest-stakes and least-rehearsed activities in the entire system. A UAT that never closed a period has not tested the thing most likely to go wrong.

Key terms

The vocabulary, briefly

Term What it means
Test script A documented scenario with steps and expected results, written from a business process.
Entry criteria What must be true before UAT starts — build complete, data loaded, environment stable.
Exit criteria What must be true before sign-off, including severity thresholds for open defects.
Severity How badly a defect affects the business, as distinct from how quickly it will be fixed.
Regression Re-testing what previously worked, because fixes break things.
What Goes Wrong

Common mistakes worth avoiding

Project team running UAT

They know the intended path and will unconsciously avoid the edges. It has to be the users.

Sample data instead of migrated data

The problems live in your real data's history and oddities.

Change requests logged as defects

A missing feature that was never in scope is a change request. Mixing the two makes the defect count meaningless.

Sign-off under schedule pressure

A signature obtained to hold a date transfers the risk to the first live month.

What to measure

Numbers worth tracking

Metric What it tells you What distorts it
Script coverage of real scenarios Whether tests reflect what the business does, including the exceptions.
First-pass rate Build quality. Very high rates early often mean scripts are too easy, not that build is excellent.
Open defects by severity at exit Against agreed criteria. Going live with open high-severity defects should be a recorded decision.
Defects found post-go-live vs in UAT Testing effectiveness, measurable only afterwards but worth knowing for next time.
FAQ

Questions we are asked about this

Long enough to run each core process with real data and to retest what was fixed. Two weeks is a common minimum for a mid-sized ERP, and it should not be the two weeks before go-live.
System testing verifies the software works as built, usually by the project team. UAT verifies the configured system supports the business, and is run by the people who will use it.
The people who will use the system daily, not the project team and not IT. They are the only ones who will notice that a screen is technically correct and operationally unusable, or that a step everyone forgot about is missing. This means releasing them from their day jobs for the duration, which is a real cost that plans frequently omit and that gets paid anyway in a worse form.
Reduce scope rather than coverage: test fewer processes properly rather than all of them superficially. Prioritise by business impact and by how often something runs — order to cash and period close before an annual report. Compressing UAT by testing everything shallowly produces a sign-off that means nothing, and the defects simply move from the test cycle to the first week of live operation, where they cost more.
Log, triage by severity and business impact, fix, retest, and regression test. The part that gets skipped under pressure is the regression, which is how a fix for one defect quietly breaks two things that previously worked. Agree the exit criteria before UAT starts — specifically, what severity of open defect is acceptable at go-live — so the decision is made calmly rather than in the final week.
No, though it is often treated as a substitute and it makes a poor one. UAT verifies the system does what the business needs; training teaches people to use it. Users participating in UAT do learn a great deal, which is a genuine benefit, but they are a small group and they are learning the system as it was during testing. Separate training, with its own materials and its own schedule before cutover, is still required.
Related

Where to read next

ERP implementation phases

Read more

ERP cutover planning

Read more

ERP data migration

Read more

Need this applied to your own systems?

We build and support the systems this page describes. Start with a free working proof of concept — no cost, no commitment.