Even when asking the same question, the response time from your chatbot can vary. However, when manually testing it’s hard to tell if long wait times is just a red herring or is a trend across all your conversations.You can view the runtime distribution across your entire Test Suite in the Analytics Platform, and averages for each specific Test. You can view your trends over time, as well as specific times after a Suite Run. This is particularly important when comparing different LLM models in your chatbot architecture, or even comparing new releases of the same model.

When reviewing reports generated in the Analytics Platform, it’s important for you to quickly understand where improvement is most needed in your chatbot. The average runtime of each Test in a Suite Run is compared to its average in a previous Suite Run (that you can configure) to help you quickly identify which Tests have large changes in their performance.

The response time of a chatbot can vary heavily depending on the number of concurrent users it needs to serve. In order to accurate achieve performance testing results, the bottest.ai Execution Engine will execute Tests with a configurable level of concurrency to mimic production conditions. Testing with high concurrency will ensure that your chatbot can handle the anticipated usage volume without delays, providing insights into the real response times your users will face in production.