Batch testing and iteration in Testing Center
Key Takeaway
Batch testing runs a whole test suite of test cases at once and scores each response automatically. Use the results to fix the agent, then test again until accuracy is acceptable.
Batch testing in Testing Center checks an agent against many test cases at once, instead of one conversation at a time. Scorers grade each response automatically, so you can see where the agent falls short before it reaches users. You then fix the agent and run the tests again, repeating until accuracy is acceptable.
What batch testing is and when to use it
In the Preview panel, you test an agent by typing one utterance at a time. Batch testing instead runs a whole test suite of test cases together, so a single run covers many scenarios. Scorers evaluate each response automatically.
The recommended order is to start in the Preview panel. Manually test a variety of inputs there and revise the agent. Then move to batch testing for broader coverage.