What to Automate?

19 09 2011

I had this interesting conversation on Automation the other day with my colleagues at my new job. It started off fairly innocently on how the automation should be thought out and what needs to be done to automate. We already have an existing framework and test scripts in place, which very efficiently work and report issues. The problem comes when we need to provide data on what automation actually does. How do you prove the effort spent on automation? You could easily say that it saves time and resources by checking for faults early in the development cycle, but how can we be sure that it actually covers scenarios that check the application? This is where a business requirement document becomes a necessity.

The issue which we increasingly face today is how do we relate the business requirements to what we test. There are a few things which are given in textbooks and across such certifications organizations like ISTQB, which provide information on such. The thing is ultimately it depends on the person who is sitting and working on the application (which most of the Program/Product Managers miss out on) and the person who is writing the scripts to automate the application testing. The best way to figure out for a restricted zone (propriety) application is to go and sit with the users and find out what they use the most (or run a key stroke capture software and see where all it goes, limitations of this later). With a public/global reach software, it is best to give out Beta versions of the same, like most of the Big Organizations do and see what is reported back.

The other end of the spectrum is propriety software, which might not have a user interface (system tools like Compilers is one example). For this the technique is basically to read through the Software Requirement Document/Specification and have confidence in your abilities to decipher the jargon written in those and convert them to simple English. I got my initial training on writing code on these only, and the golden rule for us was [in a summary]:

  • Read through the document and write what you have understood for each function
  • Create the Algorithm for each in plain English pseudo code
  • Convert these into test cases, and run those test cases
I think the above rules of going through the steps of Why, What and How has helped us out a lot to become what we are today 🙂