User Acceptance Testing That Catches Real Problems
A system can pass every test and still fail in its first busy week. Good user acceptance testing (UAT) covers the awkward cases and handovers that nobody scripted.

Plenty of new systems pass testing and then fail in their first busy week of real trading. Usually someone did test, but the testing checked whether screens worked, while the real problems sit in the awkward cases, the month-end jobs and the handovers between teams that nobody scripted.
What user acceptance testing is for
User acceptance testing, or UAT, is the stage where the people who will use the system confirm that it supports their real work. It comes after the vendor or developers have tested their own configuration. UAT asks one question: can we run our business on this?
Vendors test against the specification. Users test against reality, which includes exceptions, bad data, interruptions and the way two departments pass work between them. If UAT simply repeats the vendor’s testing, it adds very little.
Build test scenarios from real work
Good UAT scenarios come from the business, not from the system menus. Write them as short stories that run from start to finish. A strong set usually includes:
- Everyday transactions at real volume. Not one sales order but a realistic morning of them, entered by the people who will do it, at the speed they normally work.
- The awkward cases. Partial deliveries, returns after invoicing, a customer over their credit limit, a product sold both by the piece and by the carton, a price change in the middle of the month.
- Cross-department flows. An order from sales through warehouse picking, delivery, invoicing and payment receipt, with each team doing its own step in its own system role.
- Periodic jobs. Month-end close, stock counts, payroll runs, tax returns (general sales tax in Jordan, value added tax (VAT) in several Gulf states), and any e-invoicing submission the system must support.
- Permissions. Checking that a warehouse clerk cannot approve a discount, and that a branch manager sees only their own branch.
- Reports people depend on. Comparing key report figures with what the old system or the finance team would expect for the same period.
A useful source of scenarios is your own history. Ask each department for the five most annoying situations of the past year: the order that went wrong, the return nobody knew how to process, the customer who was invoiced twice. If the new system handles those, it will handle most ordinary days.
A scenario: a retailer testing new till and back-office systems
Imagine a fashion retailer with eight stores across the Gulf moving to a new point-of-sale (POS) system connected to an enterprise resource planning (ERP) system. The vendor’s testing went well: sales post, stock updates, receipts print. UAT is scheduled for one week.
If store staff only repeat simple sales, the system will pass. But the stores also handle exchanges for a different size, gift vouchers bought in one country and used in another, returns of items bought before the new system existed, and end-of-day cash reconciliation when the network drops for an hour. A useful UAT scripts each of these with a store supervisor at the till, then has finance check the postings in the ERP the next morning. Those are the cases that otherwise appear on the first weekend of trading, in front of customers.
The retailer also runs one full “day in the life” test before signing off: a single store opens, trades, handles returns and closes on the new system, with head office running the overnight jobs and checking the morning reports. Problems that only appear when steps run in sequence, such as a stock transfer that blocks the end-of-day close, show up here and nowhere else.
Run UAT as a structured process
- Agree entry criteria. UAT starts only when the vendor’s own testing is complete, data is loaded and key users are trained. Testing an unfinished system wastes users’ time and patience.
- Use migrated data, not demo data. Real customers, products and opening balances reveal problems that clean sample records hide.
- Assign named testers per scenario. Each scenario has an owner from the business who runs it and signs off the result.
- Protect their time. Testers need time released from daily work. UAT squeezed between other tasks gets done badly or not at all.
- Log every issue in one place. One list with a description, steps to reproduce, screenshots, severity and owner. Not a WhatsApp group.
- Retest fixes. Every fix goes back to the person who found the problem, and related scenarios are rerun in case the fix broke something else.
- Decide go-live against exit criteria. Agree in advance what must be true to go live, for example no open critical issues and documented workarounds for major ones.
Classify issues so decisions are clear
Not every issue should stop a go-live. A simple severity scale keeps the conversation focused.
| Severity | Meaning | Typical decision |
|---|---|---|
| Critical | Core process cannot run, or financial data is wrong | Must be fixed before go-live |
| Major | Process runs but with significant effort or risk | Fix before go-live, or agree a documented workaround |
| Minor | Inconvenience, cosmetic issue or rare case | Schedule after go-live |
| Change request | Works as specified, but users want it to work differently | Goes to change control, not the defect list |
The last category deserves attention. Many UAT logs fill up with requests for new features, which can drown out the real defects. Separating them keeps the go-live decision honest and keeps scope under control.
Testing that protects the go-live
A week of serious UAT costs far less than a month of firefighting after launch, and it builds the confidence key users need to support their colleagues. Planning test scenarios from real processes is a natural part of our Enterprise Systems work, because the people who mapped the operations already know where the awkward cases live. The best sign-off is a key user who has tried to break the system with their own worst days and could not.



