Data test
A data test is a query that looks for rows breaking a rule. The test passes when it finds none.
Data tests check the tables; test questions check the agent. A test question is one of the questions people ask most, and the agent's result is compared with an agreed answer the metric's owner confirmed.
Data tests run each time the pipelines bring in new data, and they should also run automatically on every proposed change to the data model's code, so a change that breaks a number never ships. When a test has failed, the agent should say the number is on hold, which it can do only if its connector shows it the test results.
Example
The data model filters out test orders and removes duplicates. A data test for each rule looks for rows breaking it: a test order still in the orders table, or an order counted twice. If the query finds none, the test passes; if it finds one, the return on ad spend (ROAS) built on those orders is on hold.