5

I'm a developer trying to find some guidance for the testers on our team. Our strategy regarding selenium automation testing has been rocky to this point. (A .NET project for context).

We started off with an automation test for each story, and were able to link the test to the story by the story number and the method name for the automated test. Most stories were tested, some just had stub tests where the functionality was not suitable for or did not require testing.

This resulted in a lot of code and became unmanageable.

Later we decided to scale back the automation tests to test core functionality.

What I'd like to know - is how do you track what functionality is and isn't tested so we have some confidence as developers that our code is being exercised correctly in key areas.

3
  • 4
    Please define "automation test". Are we talking Unit tests by devs, automated selenium tests or something else? thanks Jul 2, 2020 at 8:19
  • Additionally to what Durrant said, "coverage" is a multifactor attribute: sqa.stackexchange.com/a/44995/12740 What do you mean by coverage? Jul 2, 2020 at 15:45
  • @Micheal Durrant - automated tests using selenium
    – gbro3n
    Jul 6, 2020 at 19:12

4 Answers 4

6

Don't track and link stories to tests this way. It will lead to a massive mess.

Treat your automation itself as the product.
Use the Agile principle of working software (automation in this case) over comprehensive documentation such as detailed stories and links.

Reset. Separate your tests from your stories. Create a test suite.

Use test suites to validate functionality. Stories will require adjusting, changing, add new cases, etc. The test suites themselves remain as a a separate entity. There will be varieties such as smoke tests for deployments, browser and device testing, etc.

Also make sure you are educated in and promoting good practices and approaches such as:

  • Unit, Integrated, E2E and exploratory testing
  • The Agile Test Pyramid & Agile testing quadrants
  • One team approach with qa truly embedded in dev

At the end of the time day remember that tests are about adding confidence in functionality and protecting your users. Focus less on direct links to specific stories.

2
  • You're right. It did lead to a massive mess. What I'm not clear on is if we can't trace back to stories, how do we have "confidence" that a given story is properly implemented. With Selenium tests, you can't track code coverage as far as I know. I guess you can just write tests for fixes in response to problems that arise so we have confidence that that issue won't arise again without being detected?
    – gbro3n
    Jul 10, 2020 at 13:25
  • Relate tests to functionality and stories. Stories ideally like the bdd kind. Both of these are different from code Jul 10, 2020 at 13:49
5

The first question to ask yourself is if this level of traceability is necessary. For some people, maybe the answer is yes. For others, perhaps not. If you don't need traceability between story and test, you can focus on better ways to measure test coverage and correctness.

The second question to ask is how you intend to implement automation. There are different strategies, each of which would lend itself to various methods of tracking the work. For example, requiring all necessary automation to be a completion criterion for the story is different than requiring test cases to be written but automation following at a later date. The person responsible for automation, either a developer, a test specialist embedded with the team, or a separate test automation team, also drives how you trace and track the work.

Ultimately, though, I consider the "trace changes to tests" to be a very different question than "track what functionality is and isn't tested".

Test coverage is an excellent first step toward identifying what functionality is and isn't covered. Depending on your tools, you may be able to combine coverage from different levels of tests to see what is exercised by unit tests, integration tests, and acceptance tests. Using coverage, you can identify what parts of the system need more tests.

However, coverage doesn't say anything about how good your tests are. Just because they cause lines of code to be executed doesn't mean they assert anything valuable about the code or its intended effects on the system. Peer reviews of tests and test suites can help with this. Mutation testing can also help provide insights into the ability of your tests to detect changes in the software under test and fail.

How you monitor coverage, carry out peer reviews, and otherwise assess the quality of your tests depends on who is responsible for creating and maintaining the test suite.

4
  • Thanks for an great answer ! Could you please elaborate more how mutation testing can help in gaining confidence in tests? Jul 5, 2020 at 9:31
  • @VishalAggarwal A good answer to that won't fit in a comment and it would just add noise to this answer that's not directly relevant to this question. I'd recommend doing some research on mutation testing and test quality and, if you still have questions, asking a new question about it. Maybe I'll write an answer there... Jul 5, 2020 at 10:19
  • +1 for mutation testing. I was not aware of this. @Vishal Aggarwal - it makes modifications to application code (or mutants) and verifies whether your unit tests catch the mutations: opensource.com/article/19/9/mutation-testing-example-definition
    – gbro3n
    Jul 10, 2020 at 13:41
  • @Thomas Owens , The reason I asked for detail(in the context only) as you brought it up as part of your answer in the context of this question. :) Jul 11, 2020 at 14:42
1

Use 'Tagging' feature of defect tracking systems like Jira/TFS to add tags like 'automated' to stories and use reporting mechanism based on 'tagging' to get desired reports of coverage assuming there is no issue with the quality of tests themselves.

Also would suggest to not to try covering each small validation in automation but focus only on the happy path to avoid unnecessary maintenance overhead.

0

For reporting purposes: I find it useful to link everything in a Requirements Traceability Matrix. It gives a quick glance of how use cases get broken down into requirements, which get broken down into features (tracked by development). You can just add an additional column for automated tests + number of tests implemented. If the cell is not checked it means that the feature requires additional testing and cannot be regressed just running the automated test suites.

For work planning and tracking purposes: How you structure the tests / tests suites is up to you. The tests can be planned at the same time as other sprint activities, but the work should be tracked using their own user stories.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.