Blog post

Paths to Automation

By Thomas Murphy | June 30, 2016 | 2 Comments

Digital business creates a strong drive for companies that haven’t traditionally thought of themselves as software companies to become more like software companies. One strong need continues to be how to drive greater amounts of test automation.  We see two primary paths that have emerged and as we work on updating our Magic Quadrant for Software Test Automation this creates fundamental choices in tools.

Option 1: Become more like a software company or your favorite internet devops prototype. In this case your path is to use developers or testers with development skills to build tests.  Typically these tests will be built using tools like Selenium and associated frameworks. The advantages are Se is open source gaining a lot of community support, it connects into the devops toolchain readily including storage of the tests in the same repository that the code is stored in. The disadvantage is you are writing code to test your code meaning tester need stronger development skills and it can be less efficient. This generally means hiring new testers with these skills the Software Development Test Engineer that is being frequently sought after. 

Option 2: Is more of a move forward with the people and skills we have. This means I need to find ways to make non-developer testers effective at automation.  This has been a long running challenge leading to a number of often failed attempts.  Record and playback is a great example. It demos well and you can get automation but it won’t be maintainable and doesn’t easily translate to a good understanding of the quality of the test suite.  For most organizations, outside of wholly new business units (e.g. building a new mobile app team) this is the reality we have potentially hundreds of employed testers who have solve business process knowledge but for which an approach to automation is raw coding/scripting is not viable.  Here is where we see growing importance of frameowkrks, model based approaches and machine learning playing a critical roll.

We would love to hear your thoughts and success/failure stories either here or on the Gartner Client discussion portal. The need to automate is becoming critical please join the conversation.

Leave a Comment


  • Hi Tom,
    Great topics. As you might imagine, we are seeing far more organizations moving toward Option 2, especially in the bi-modal enterprises. While Option 1 works well in smaller project teams, or pure web shops, the enterprise cannot support the skills needed across diverse mode 1 and 2 applications. This is our primary client, who is automating end to end with “manual” testers becoming automation specialist without increasing developer skill sets.

  • Brad Matsugu says:

    Very relevant to our conversations we’re having as well. There is a lot of confusion and we are finding a mix of both Option 1 and 2 as companies are still finding their way. Option 1 has become popular as developers and testers are being involved in the planning process much earlier. Option 2 as Tom mentioned has been tried and failed but the emergence of new tools has allowed non-developers to produce tests with limited technical background.