Change takes place rapidly in Agile projects. This change means that test status, test progress, and product quality constantly evolve, and testers must devise ways to get that information to the team so that they can make decisions to stay on track for successful completion of each iteration. In addition, change can affect existing features from previous iterations. Therefore, manual and automated tests must be updated to deal effectively with regression risk.
Communicating Test Status, Progress, and Product Quality
Agile teams progress by having working software at the end of each iteration. To determine when the team will have working software, they need to monitor the progress of all work items in the iteration and release. Testers in Agile teams utilize various methods to record test progress and status, including test automation results, progression of test tasks and stories on the Agile task board, and burndown charts showing the team’s progress. These can then be communicated to the rest of the team using media such as wiki dashboards and dashboard-style emails, as well as verbally during stand-up meetings. Agile teams may use tools that automatically generate status reports based on test results and task progress, which in turn update wiki-style dashboards and emails. This method of communication also gathers metrics from the testing process, which can be used in process improvement. Communicating test status in such an automated manner also frees testers’ time to focus on designing and executing more test cases.
Teams may use burndown charts to track progress across the entire release and within each iteration. A burndown chart [Crispin08] represents the amount of work left to be done against time allocated to the release or iteration.
To provide an instant, detailed visual representation of the whole team’s current status, including the status of testing, teams may use Agiletask boards. The story cards, development tasks, test tasks, and other tasks created during iteration planning are captured on the task board, often using color-coordinated cards to determine the task type. During the iteration, progress is managed via the movement of these tasks across the task board into columns such as to do, work in progress, verify, and done. Agile teams may use tools to maintain their story cards and Agiletask boards, which can automate dashboards and status updates.
Testing tasks on the task board relate to the acceptance criteria defined for the user stories. As test automation scripts, manual tests, and exploratory tests for a test task achieve a passing status, the task moves into the done column of the task board. The whole team reviews the status of the task board regularly, often during the daily stand-up meetings, to ensure tasks are moving across the board at an acceptable rate. If any tasks (including testing tasks) are not moving or are moving too slowly, the team reviews and addresses any issues that may be blocking the progress of those tasks.
The daily stand-up meeting includes all members of the Agileteamincluding testers. At this meeting, they communicate their current status. The agenda for each member is [Agile Alliance Guide]:
- What have you completed since the last meeting?
- What do you plan to complete by the next meeting?
- What is getting in your way?
Any issues that may block test progress are communicated during the daily stand-up meetings, so the whole team is aware of the issues and can resolve them accordingly.
To improve the overall product quality, many Agile teams perform customer satisfaction surveysto receive feedback on whether the product meets customer expectations. Teams may use other metrics similar to those captured in traditional development methodologies, such as test pass/fail rates, defect discovery rates, confirmation and regression test results, defect density, defects found and fixed, requirements coverage, risk coverage, code coverage, and code churn to improve the product quality.
As with any lifecycle, the metrics captured and reported should be relevant and aid decision-making. Metrics should not be used to reward, punish, or isolate any team members.
Managing Regression Risk with Evolving Manual and Automated Test Cases
In an Agile project, as each iteration completes, the product grows. Therefore, the scope of testing also increases. Along with testing the code changes made in the current iteration, testers also need to verify no regression has been introduced on features that were developed and tested in previous iterations. The risk of introducing regression in Agile development is high due to extensive code churn (lines of code added, modified, or deleted from one version to another). Since responding to change is a key Agileprinciple, changes can also be made to previously delivered features to meet business needs. In order to maintain velocity without incurring a large amount of technical debt, it is critical that teams invest in test automation at all test levels as early as possible. It is also critical that all test assets such as automated tests, manual test cases, test data, and other testing artifacts are kept up-to-date with each iteration. It is highly recommended that all test assets be maintained in a configuration management tool in order to enable version control, to ensure ease of access by all team members, and to support making changes as required due to changing functionality while still preserving the historic information of the test assets.
Because complete repetition of all tests is seldom possible, especially in tight-timeline Agile projects, testers need to allocate time in each iteration to review manual and automated test cases from previous and current iterations to select test cases that may be candidates for the regression test suite, and to retire test cases that are no longer relevant. Tests written in earlier iterations to verify specific features may have little value in later iterations due to feature changes or new features which alter the way those earlier features behave.
While reviewing test cases, testers should consider suitability for automation. The team needs to automate as many tests as possible from previous and current iterations. This allows automated regression tests to reduce regression risk with less effort than manual regression testing would require. This reduced regression test effort frees the testers to more thoroughly test new features and functions in the current iteration.
It is critical that testers have the ability to quickly identify and update test cases from previous iterations and/or releases that are affected by the changes made in the current iteration. Defining how the team designs, writes, and stores test cases should occur during release planning. Good practices for test design and implementation need to be adopted early and applied consistently. The shorter timeframes for testing and the constant change in each iteration will increase the impact of poor test design and implementation practices.
Use of test automation, at all test levels, allows Agile teams to provide rapid feedback on product quality. Well-written automated tests provide a living document of system functionality [Crispin08]. By checking the automated tests and their corresponding test results into the configuration management system, aligned with the versioning of the product builds, Agile teams can review the functionality tested and the test results for any given build at any given point in time.
Automated unit tests are run before source code is checked into the mainline of the configuration management system to ensure the code changes do not break the software build. To reduce build breaks, which can slow down the progress of the whole team, code should not be checked in unless all automated unit tests pass. Automated unit test results provide immediate feedback on code and build quality, but not on product quality.
Automated acceptance tests are run regularly as part of the continuous integration full system build. These tests are run against a complete system build at least daily but are generally not run with each code check-in as they take longer to run than automated unit tests and could slow down code check-ins. The test results from automated acceptance tests provide feedback on product quality with respect to regression since the last build, but they do not provide status of overall product quality.
Automated tests can be run continuously against the system. An initial subset of automated tests to cover critical system functionality and integration points should be created immediately after a new build is deployed into the test environment. These tests are commonly known as build verification tests. Results from the build verification tests will provide instant feedback on the software after deployment, so teams don’t waste time testing an unstable build. Automated tests contained in the regression test set are generally run as part of the daily main build in the continuous integration environment, and again when a new build is deployed into the test environment. As soon as an automated regression test fails, the team stops and investigates the reasons for the failing test. The test may have failed due to legitimate functional changes in the current iteration, in which case the test and/or user story may need to be updated to reflect the new acceptance criteria. Alternatively, the test may need to be retired if another test has been built to cover the changes. However, if the test failed due to a defect, it is a good practice for the team to fix the defect prior to progressing with new features.
In addition to test automation, the following testing tasks may also be automated:
- Test data generation
- Loading test data into systems
- Deployment of builds into the test environments
- Restoration of a test environment (e.g., the database or website data files) to a baseline
- Comparison of data outputs
Automation of these tasks reduces the overhead and allows the team to spend time developing and testing new features.