Write app requirements around a real task, specifying who performs it, what the system must record and how exceptions should be handled. Use those outcomes as acceptance criteria that frontline staff and back-office colleagues can test together. “Add goods receiving” leaves important decisions open: who can confirm the receipt, how a short delivery is handled and when the stock balance changes. Leaving those decisions until the first demonstration can mean reworking the interface, data model and reporting.
Describe the result before designing the screen
For each requirement, name the user, the information they need and the record that should exist when the task ends. In a hypothetical warehouse, a worker can open a purchase order assigned to their warehouse and enter the quantities received. The inventory supervisor can see the same receipt in the back office. If a quantity needs correcting after submission, only an authorised person can approve the amendment, and the original action remains traceable.
Write expected outcomes for common exceptions
- A partial delivery records stock received and leaves the outstanding quantity available for a later receipt.
- An over-delivery shows the discrepancy and routes it to a supervisor under the existing approval policy.
- A connection drop after submission leaves a clear pending state. Retrying checks for a previous receipt before creating another.
- A user assigned to another warehouse cannot see or change this purchase order.
Include one case where an operation looks successful on the phone but fails at the system of record. Have someone check both sides before passing the case. A saved screen is only part of the job if finance or inventory relies on the record downstream.
Give testers the same data and version
Keep a repeatable test set with fictional staff accounts, purchase orders and expected amounts. Each case should identify the role, test build, steps, expected result and evidence. Ask the warehouse team to check the flow and the supervisor to reconcile the changes in the back office. Agree which defects block release and which can be scheduled later. Name the business owner who signs off, and give the owner evidence rather than a simple pass total.
Run a small pilot before wider release
Apple TestFlight and Google Play testing tracks allow teams to distribute test builds to selected users. Ask testers to report their build, device and the steps they took, with a non-sensitive example record where useful. After sign-off, update training material and make someone responsible for the first round of support. That creates a way to catch differences between controlled test data and a normal working day.
Frequently asked questions
Do we need finished designs before writing requirements?
Start with outcomes, responsibilities and exceptions. A simple workflow or prototype can then clarify how the screens should support them.
Which problems should block release?
Prioritise anything that prevents core work, corrupts important records or permits unauthorised access. Have the business owner agree how remaining defects will be handled.
Can acceptance criteria change during development?
Yes. Record why the change is needed, which tests it affects and any consequences for scope or timing. Share that revision with everyone who must use it.

