What makes a good automated test?

I have been asked to try automate the testing of a wider range of for the group. This is going to be quite challenging since I have not automated desktop applications for about 15 months and the software we make is quite unique. I will also hopefully be teaching the other testers how to maintain the automated tests and eventually create their own.

This challenge has made me think back on my automated testing experience and ask "What makes a good automated test?". A good automated test has all the hallmarks of a good manual test except that its comparatively quick and cheap to run. This means that an automated test is successful when it finds bugs and repeatable where ever the software is installed.

In terms of web testing it would be a test script that can be created and then stored in a central place and be run from the central place. Ideally this would be also run on the continuous integration server when doing builds.

The thing that makes a good automated test great is scalability. I recently watched a YouTube video of the Selenium Users Meet-up. There were Google staff there saying that their testing farm handles 51000 tests and most Friday's all of these tests run. They then went on to complain that Selenium struggled to scale to that size. To be honest at that size I am not surprised!




P.S. To those who are waiting for the next Selenium Tutorial I am hoping to have it complete in the next week for everyone to use.