# Smart Test Selection Explained ## How does Appsurify's Model determine Test Selection Simple answer, based on recent developer changes and the trained model! When Appsurify connects the Repository and begins to receive Incoming Commit data - Appsurify initially analyzes the Repository for basic structuring. Once the Automated Tests are connected, Appsurify leverages Patent-Pending proprietary AI-Risk Based Testing Technology to train each company's **unique AI-Model** based off of incoming commits and the corresponding test results. Over a period of 2-3 weeks, Appsurify trains a robust AI Model that links **Code to Tests** - because it will know which Commits impacted which Tests to form a union between the two. Once the AI Model is Trained, when a new Commit comes in, Appsurify will know which Tests are associated with that region of the Codebase - and subsequently; select and execute the relevant tests by order of Priority automatically in the CI/CD based on the Parameters set by the team The AI Model is recalibrated on a rolling basis to ensure it's always up to date. For more information on on our Test Selection please see our [Test Selection Whitepaper](https://appsurify.com/resources/test-selection-with-appsurifys-testbrain/) on the Appsurify website. ### **Will Appsurify catch every Failed Test?** **Appsurify is designed to catch as many defects as possible by running as few tests as possible.** Appsurify is designed to be efficient in finding bugs in its Smart Subset while leaving room to find other potential defects that cause other test failures. \ \ 1 Bug often causes more than 1 Test Failure, often multiple. To be efficient and fast, Appsurify is designed to catch as many bugs in as few tests as possible. {% hint style="info" %} **Example:** * 1000 Tests. * 50 Bugs. * 200 Tests Fail because of the 50 bugs. * Appsurify: 10% Smart Subsets \ If Appsurify is enabled and running the **Top "10%"** of Tests, it will be forced to Select and Execute what the Model deems to be the most important 100 Tests most likely to Fail. \ Therefore, Appsurify will Prioritize and Run the Tests most likely to Catch the underlying Defects. \ \ **In this case**, Appsurify will run Smart Subset of **100 Tests**. In that example subset, it could be the case that 70 of them are Failed Tests and 30 of them are Passing Tests. The **IMPORTANT** thing to note is that the 70 Failed tests run cover the majority of those 50 underlying bugs and Raise them as New or Reopened Defects. \ \ As is often the case with a single defect, it will likely cause multiple Failed Tests. \ \ Therefore, Appsurify needs to execute only one or a portion of those Failed Tests in order to raise the Defect while saving room in the Smart Subset to catch other defects. \ \ To further this case, if the 50 Bugs caused 50 Tests to Fail **(1-to-1)**, if Appsurify's Smart Subset of 100 Tests caught 50 Failed Tests and 50 Passed Tests and still caught and Raised all 50 Bugs - that would be very successful.\ \ Appsurify is designed to find the underlying defect by running as few tests as possible. \ \ If the user decided to change the Smart Subset to **"5%"** in this case, Appsurify would still have room in the Smart Subset to find all 50 Bugs by only running 50 Tests. \
{% endhint %}