There are three ways you can use to test a rule set: unit test, object test, and regression test.
- Unit test: when doing unit test, it shows which value to provide. After hitting the “Run test”, we can see the rule’s results.
-
Object test: when doing object test, you need to select a data object and then run a test.
-
Regression test: use this to make sure a rule set keeps producing the expected results as you keep changing it over time. Open the rule set's Testing window and go to the Regression tab.
Why use regression testing
Unit and object tests only tell you what the rule set does right now. They don't warn you if a later change accidentally alters an outcome someone was relying on.
Regression tests solve that: you save a test case once (an input and what you expect it to return), and from then on you can re-run all of your saved cases with one click to confirm nothing has changed. This is especially useful before publishing a new release, or after several people have made changes to the same rule set.
Creating a regression test
-
Open the rule set and go to Testing → Regression.
-
Click New test. If the rule set has no tests yet, the form is already open.
-
Choose Data object to pick a saved data object, or JSON to type the input. Picking a data object fills the Title with its name. You can change the title.
-
Under What to check, choose what the test should expect:
- Everything it returns now is available when you use a data object. When you save the test, the rule set runs on that input and the whole result becomes the expected result.
- Specific fields: click Add field, then choose a field, an operator, and the expected value. Available operators include equals, not equals, greater than, less than, greater than or equal to, less than or equal to, contains, is null, and is not null.
You can use both options in one test.
-
Click Save test to save it, or Cancel to close the form.
Repeat this for every scenario you care about (e.g. one test per customer segment, edge case, or previously reported bug).
Save a run as a regression test
Under the results of a unit test or an object test, click Save as regression test. This saves that run's input and output, with the output as the test's expected result. It's a quick way to save a result you have just checked by hand.
Running regression tests
Saved tests appear first in the Regression tab. Above them are Run all, a count such as “3 tests · 2 pass · 1 changed” (errors are counted with fails), and New test. Statuses with no tests are not included in the count.
Each test appears in a row with its status, title, what it checks, and when it last ran. What it checks is either “everything it returns” or a field check, such as whether the decision equals APPROVED. Click the play button to run that test. The menu at the end of the row has Edit and Delete. Delete asks you to confirm before removing the test.
- Run a single test: click the play button on that test's row to check it in isolation.
- Run all: click Run all at the top of the tab to run every saved test.
After Run all, each test that is Changed or Fail opens by itself and shows a table with Field, Expected, and Got. A test that checks everything it returns also shows how many of its fields still match. Once a test has run, you can open it with the arrow at the start of its row.
Tests have one of these statuses:
| Status | Meaning |
|---|---|
| Pass | The rule set produced what the test expected. |
| Changed | The result differs from what the test expected. |
| Fail | A specific field check did not match its expected value. |
| Error | The test could not run at all, for example because its input could not be read. |
| Never run | The test has not been run. |
A test that comes back Changed or Fail is not necessarily a bug. The rule set may have been updated on purpose, and the test's expectation may need to be refreshed. If the new behavior is correct, open the test with Edit, tick Everything it returns now again, and click Update test. The current result becomes what the test expects.
Recommended approach
- Before releasing changes to a rule set, run Run all in the Regression tab and confirm nothing unexpected shows as Changed or Fail.
- When you fix a reported issue, add a regression test that reproduces it, with a specific field check for the corrected result, so the same mistake can't silently reappear later.
- Keep test names descriptive (e.g. "high income, poor credit → manual review") so a Changed or Fail result is easy to interpret without re-reading the input.