What tests run against.
WHAT GOOD TEST DATA IS
Realistic in shape and volume Covering the cases that matter Reproducible Free of real personal information
WHAT THE APPROACHES ARE
Generated data Data created by the test itself Anonymised production data Curated fixed datasets
WHAT TESTS CREATING THEIR OWN DATA PROVIDE
Independence and reliability.
WHY THAT IS PREFERABLE FOR AUTOMATED TESTS
They do not depend on state anyone else can change.
WHAT SHARED FIXED DATASETS PROVIDE
Consistency, and realistic complexity.
WHAT THEY COST
Tests coupled to them, breaking when they change.
WHAT PRODUCTION DATA PROVIDES
Realism nothing else matches.
WHAT IT REQUIRES
Anonymisation, thoroughly.
WHY THOROUGHLY
Names removed while addresses, dates and relationships remain is not anonymised.
WHAT TO REPLACE
Every identifying field, consistently, so relationships survive.
WHAT TO NEVER DO
Copy production data to a test environment without anonymising.
WHY
Test environments are less protected, and this is a common cause of breaches.
WHAT TO GENERATE
Volume, where realistic size matters.
WHAT TO MAINTAIN
The data alongside the tests.
