Data independence

Twelve Contoso testers start UAT on Monday morning. By 10:00, three of them are reporting defects that are not defects. One tester's order was delivered by another tester. A customer that should pass the credit check is blocked because another scenario used up its limit. Stock that should be there was issued by a colleague in a different scenario. This is what shared test data does.

How shared data breaks tests

Each of these produces a false failure (a "defect" that is only a data clash) or, worse, a false pass. False failures waste the defect team's time and erode trust in testing. False passes hide real problems. Both are avoidable.

Ways to keep data independent

• Dedicated records per tester or per scenario. Each tester gets their own customers, suppliers and purchase orders. Create them in a batch before the cycle starts.
• Tagging. Mark test records so everyone can see whose they are. For business partners, a search term such as UAT-O2C-MARIA works well and is easy to filter.
• Separate stock buckets. Use separate storage locations, or separate products that behave identically, when several testers need stock of the same item. Contoso gives O2C testers 171A and production testers 171B.
• Never change shared master data during a test. If a scenario needs a different product setting, create a dedicated product for it.
• Consume-once data. Treat data that a scenario uses up (stock, open orders, open items) as consumed and plan enough for retests.

Copying a reference record

The fastest way to create many similar records is to build one good reference record and copy it. For business partners, the customer role data, sales area data and tax classification come along; for products, the views and plant data. Check what the copy does not carry over (condition records, credit data and stock are separate objects) and add those to the setup script. Keep the reference record itself read-only, and never use it in a scenario, so that every copy starts from the same known state.

If your project uses internal number assignment, you will not control the numbers of new business partners. That is fine. Rely on the search term tag and record the generated number in the data sheet as soon as the record is created.

When data has to be shared

Some data cannot be duplicated: there is one company code, one controlling area, one set of G/L accounts. Integration scenarios also deliberately share data, because one team's output is the next team's input. In those cases, agree the sequence and ownership explicitly: who creates the record, who consumes it, and when. That is the subject of the integration testing course.