Agile testing – some thoughts after an excellent Elevate-evening

Posted by Marcus Hammarberg on November 13, 2009
Stats

Yesterday was another Elevate-evening and again I was impressed by the diversity and knowledge of the consultants of Avega.

The theme for last night was; Agile Testing. This is an area that has confused and frustrated (is that a word?) me for some time. I haven’t got it to work in any of the agile teams I have been leading. Here are sample of my failures (as in failure is good – an opportunities to learn :));

  • No testers and no testing in the team. This was a disaster. We did agile development but when we were done a 3 month testing phase took place. Sounds a lot like waterfall to me. And we didn’t harvest any of the goodness that agile can bring.
  • Testers in the team – but not doing agile testing. So we decided to move testers into the team. But they had to keep up with the company testing policies. So each time we did a change, how ever minor it was, it resulted in several days work for the testers; updating test cases, use cases and tests plan etc.
    That didn’t work very well either.
  • Finally we had testers in the team that only tested the things we worked with during the sprint. And with an acceptance test-phase after each 4-5 sprints. This was the best result we’ve got at this customer, but it was very hard to get through to the testers at first, and to the developers after a while…
    We got stressed and strained towards the end of each sprint, since the testers had nothing to do in the beginning and to much to do at the end. And the regression testing were killing them – more and more for each sprint.

So I very much looked forward to the evening last night since I didn’t think that any of the approaches I’ve tried was good. And quite frankly I haven’t heard anyone say; “Agile testing or testing in an agile project? Yeah I know how to get that to work smoothly”.
Finally I would get to hear how to do it!

OK – that was my introduction. Here is my thoughts after the lightning talks last night. Unfortunally I didn’t have time to stay for the Open Space.

As I hoped they all had some similar thoughts and underlying themes, which I takes as points worth noticing. For me three things stood out:

  • Team commitment to test - make the whole team think test. Make sure that you don’t have developers and tester – only team members. Which fit very nicely with the next point;
  • Definition of done – include testing in your definition of done. Or in other words; “We’re not done until the thing is tested.”
    We even got some suggestions on what to do if developers don’t take interest in testing; let testers be “sick” for a sprint, which will make developers to do the tests…
  • Invest in automation – all the success stories from yesterday included automation of regression tests. Learn some tools for that, for example Selenium and let the testers do test automation, rather than only manual testing.
    Manage the test-scripts for automation of regression tests as any other code.

All these will be in place if you can get the team to buy in on a definition of done such as:

“A story is done when it has an passing automated regression test”

Of course it will probably contain some other stuff also but only to get a feel for it.

Doing automated regression tests will have plenty of benefits;

  • the system needs to be testable – can we write tests against a service? do we need a test client that enable us to do automated tests?
  • you can start writing the regression tests before the code is done, in many cases
  • you get a suite of tests that run through the whole system and checks for breaks. Included in your Daily Build or Continuous Integration this will give you a nice status of your project.

OK – it was a great evening. Thank you everyone – you got me thinking a lot.



Published by Marcus Hammarberg on Last updated