Monday, July 27, 2009

QA Feature owner - when testers become project managers

In our R&D organization, new features are usually led by developers. While managing the new feature, the leading developer is assigned the title Feature Owner for that feature. She delegates some work to other developers while performing some coding tasks personally. The larger the feature, more time is spent on managing and coordinating efforts and in smaller ones the FO spends more time coding. The interesting issue (for me) is the interaction with the testing team.

Most features are multi tiered: there's a backend layer, some application layer and a UI. Tests for each layer are different, of course: Greybox testing for backend layer, SQL scripts for database related tests, UI Automation and manual tests for the UI, etc'. The problem is, most testers don't feel equally comfortable with each testing methodology so what tester do I sign for each feature? If I assign a manual tester to a "mostly UI" feature does it mean we want test the backend? My coder can't handle the UI automation suite, maybe I'll just skip automatic UI test?

The other alternative is assigning several testers for each feature, based on their skills. This ensures that each tester is assigned tasks she's good at but it also poses other problems:
  • The overhead of testers learning new features is now multiplied by number of testers involved in testing the feature
  • Dispatching work between the testers and writing a test plan can not be done by the development FO, as she does not have the testing know how. Also, who is to know all the bugs related to the feature? Each assigned tester? Too messy and getting a quality overview is hard.
  • Why not assign the QA team leader for dispatching tasks and writing a test plan? Because most of the time she will not do the actual testing on the feature, so we have a paradox: the QA person who studied the feature and designed the tests does not test it and the people who test it did not study it thoroughly (overhead, remember?)
The solution we are practicing now in my testing team is assigning a QA Feature Owner (QFO) to the development Feature Owner (FO). The assigned QFO is usually the person who's skills are the most relevant to the feature, but she needn't posses all the required testing skills: the QFO will assign tasks to other testers in the team in areas in which she lacks or simply does not have time to run or design tests.

This is how it works: upon reading the spec, the QFO designs a test plan, a document detailing the general testing strategy. In this document risks are identified and then tests are written as a sketch: a few words describing the general purpose of each test. We find it convenient to start by designing Greybox (GB) tests, then Automatic UI tests (AUI) and finally manual tests.

It doesn't matter if the QFO is a programmer or not when designing GB tests. Using Data Driven Testing methodologies most GB test cases are list of methods and what they are supposed to do and attached CSV files which are no more than input/output files based on the original spec. AUI tests are simply stories in English ("go here, click there") which can also be augmented beforehand using CSV files. Test cases are later reviewed by other team members to validate their testing logic and enrich the tests in areas in which the QFO is weaker. For example, a manual tester can review the manual tests written by a QFO who has strong programming background and suggest additional tests. Test cases will later be examined by the FO and QFO together, in order to ensure full coverage.

Once all test cases are approved by the QFO and feature and test coding begins, the QFO starts following the testing project. The QFO is responsible for ensuring that tests are run but not necessarily by the QFO personally: testing tasks can be assigned to other testers as long as the QFO remains the testing team's focal point for the feature. The QFO does not need to be the one reporting all defects but must know all defects opened which are relevant to her feature. The QFO is responsible for the testing of the feature but the actual tests may be performed by others.

We have started implementing this methodology only a few weeks ago and it remains to be seen whether we benefit from it on the long run. I don't take the fact that we're asking hands-on testers to become project managers for granted and believe some will adapt to it better than others. Others will resent the time spent in meetings and reporting and will feel more at home getting their hands dirty with testing. I have no intention of forcing any tester to become a QFO, I consider it a perk, not a requirement. I do, however, believe that tester assuming the role of QFO will enjoy the challenge of managing a testing project, interacting with more team members from the R&D than they are used to and experience new testing methodologies. I believe a good QFO is a potential testing leader.