Run and read tests

Every version in Ptally carries its own tests. A test is a complete set of inputs with the result you expect. Running the tests after every change shows straight away whether prices have moved.

Run all tests

Open a version and find the Tests section. Click Run all tests. Each test is priced against that version's model and the table fills in:

• Result, the price the model returned;
• Valid, whether the quote was valid or declined;
• Errors, a count of any warnings from the engine, such as a lookup that found no matching row;
• View, to open the test.

The counts at the top show how many tests returned valid and not valid quotes. Clear resets the results.

In this example, the two decline tests return Not valid, which is what they are designed to do. The cat with a dog breed also shows one warning, because its breed lookup found no row.

Check results against expectations

Open a test with View to see its name, key, input JSON, Expected result and Expected valid value. Compare the expected values with the Result and Valid columns.

If you connect an AI client over MCP, the run_version_tests tool runs every test and reports which ones match their expected result and which do not.

Add and download tests

In a draft version you can also:

• Add test, to add a single test with its inputs and expected result;
• Add batch tests, to add several tests at once.

Download tests saves the test suite as a file, which is useful for reviews and records.

Good tests to keep

• The worked examples from your rating guide or filing.
• Values either side of every band boundary.
• At least one case for each decline and referral rule.
• Every pricing bug you have fixed, so it cannot return.

For more on building a test suite, read Every rate change needs a test suite.