What criteria do you use to decide which tests to keep for manual testing and which to automate?

Start with an analysis based on the testing pyramid (Unit, Integration, End-to-End). Automation is effective at lower levels (high speed, less fragility). E2E tests are automated selectively, as they are slower and more fragile.

Automation criteria:

• Frequency of execution: Tests that are run often (regression, smoke) are beneficial to automate to save time and reduce the risk of errors due to routine.
• Stability of functionality: Functionality that is not planned to change radically in the near future. Frequent changes in interface or logic may require constant reworking of automated tests.
• Complexity and repeatability: Complex, monotonous sequences of actions that are easy to miss or perform incorrectly manually.
• Volume of data: Tests that require verification on a large set of input data.
• Execution time: Operations that take a long time to perform manually.

Criteria for leaving tests manual:

• Exploratory testing: Requires a creative approach, intuition, and unpredictable scenarios. Difficult to automate.
• Usability testing: Evaluation of user experience, interface perception, which is subjective and not easily formalized.
• Look and Feel testing: Checking pixel-perfect layout, color schemes, fonts. Automation of such checks is often fragile and requires complex logic.
• One-time functionality testing: Functionality that will be used rarely or is temporary. The costs of automation may outweigh the benefits.
• Functionality with frequent changes: If functionality is constantly being refined and changed, automation may become ineffective due to the need for constant maintenance of automated tests.
• Tests requiring human perception confirmation: For example, checking the quality of images, videos, sound.

Decision-making also depends on available resources (time, budget, team qualification).