Overview

Testing is crucial for the products that you build and that your customers use.

In the same way that you demand high-quality code from an architectural and strict development point-of-view, your test cases should also have the same high-quality.

As you struggle to make releases more often, while maintaining quality as a priority, you struggle with limited resources and limited time.

Thus, you need to make choices in order to perform the right testing with the time that you have, while maintaining a certain level of confidence.

In this document you can find some guidelines for enabling great testing, which, as we'll see, is not limited to writing test cases.

Enabling great testing

Testing quality does not just depend on the executed Tests and bugs found during that process. Testing is a lot more.

But to have great testing, you'll need first of all to have a testable SUT, which is properly described and detailed in clear, well-defined requirements, and reviewed from the start by great testers whose skills helped build a better product through awesome testing.

A testable SUT

In order for a system to be testable, it must provide certain characteristics (i.e. it must be built in such a way that promoves them) including:

• observability: test results must be easily observable
• controlability:  ability to control the state of the SUT
• automatability: provides mechanisms that allow automation to be implemented
• stability & availability: must be available and stable enough so testing can be performed efficiently in a reliable manner
• simplicity & comprehension: must be understandable, so that testing can be performed

Clear, concise and correct requirements

In order to specify great Tests for existing requirements, first of all they need also to meet certain quality criteria, by incorporating some key characteristics:

• consistency: internally and externally; by using consistent terminology and by making sure requirements are non-contradictory internally and between each other
• complete and concise, yet short: requirement scope must be well defined and limited, so that it can be implemented correctly; some caution should be taken in order to avoid, hard to digest descriptions
• correct: addresses correctly the business/stakeholders needs
• unambiguous: clear, not subject to different interpretations
• quantitative (not qualitative): quantitative aspects can be measured, while qualitative ones are ambiguous
• verifiable: a requirement must provide ways so that it can be verified; having acceptance criteria is a common practice to address this

Being a great tester

To perform valuable testing, a tester needs many skills. Some of these include the ability to:

• know the SUT, its components and the communication/interactions between them as much as possible
• understand the use cases
• understand the value proposition (what does the SUT try to solve for the end users/customers)
• have some level of understanding of the internal architecture of the SUT
• be part of the team; work together with other team members, including developers, BA, etc
• understand what is critical for the customer (using feedback mechanisms, such as metrics, support among other)

A great Test

What makes a Test, a great Test? 

There are a set of characteristics that all scripted tests should have. The following ones are just a few of them:

• goal oriented: focused on one goal, one purpose, gives meaning/value to the Test itself
  
• simple & short: simple Tests are easier to understand and to execute
• consistent: in the terminology being used, non-contradictory
• reliable & deterministic: how consistent test results are between different runs, executed under the same conditions
• objective: how clear the test is so that different testers achieve the same results
• reviewed: by having the feedback of others, your Test will become better
• valid: to what extent will test results address the purpose of the test, accordingly with the acceptance criteria defined for the requirement
• helpful on RCA (root cause analysis): helps not just on identifying a problem but on understanding the root cause
• maintainable: how easy is it to update changes in the related requirement
• discoverable: a Test should be easy to find, by multiple criteria, whenever you need it 
• fast: a Test should be executed as fast as possible as a means to reduce effort; this is specially crucial for automated testing where lengthy Tests can burden the CI/CD process
• independent: Tests that are self-contained are easier to manage
• parallel: can be run on parallel; this is specially important for automated testing in order to reduce the overall execution time

Whenever performing other types of testing, such as exploratory testing, some of the former do not need to be considered due to the nature of the practice, including objectivity and consistency among other.

Tests in Xray

Xray provides, by default, three different "Test Types" (you may create additional ones).

• Manual: a scripted test composed of steps; mainly used for manual testing but it can also be used as an abstraction of an automated test. This test kind of this test is "structured".
• Cucumber: a cucumber Scenario/Scenario Outline, normally executed in automated context (although it can also be executed manually). This test kind of this test is "gherkin".
• Generic: an abstraction of some Test that is neither Manual nor Cucumber; normally, corresponds to an automated test implemented in some framework (e.g. JUnit, TestNG, Robot, custom framework). It can also be used as a means to abstract an exploratory test/session. This test kind of this test is "unstructured".

However, one consequence of the above is that there is no explicit way to identify whether a Test is automated or not. However, you may define your own rules for that.

Best Practices

Process

Testers are part of the team

Involve testers whenever reviewing and discussing requirements or user stories with the development team; make testers part of the team and shift your testing to the left side of your development life cycle.

Requirements/stories should be clear and have acceptance criteria

Clear requirements are essential to have the correct understanding of what aims to be addressed by the requirement and the related business needs. 

Make sure they are reviewed; you can use Jira's workflow status in the related issue for this.

Test using the whole Pyramid

Have tests for the different layers of the Test Pyramid, starting with unit tests up to E2E/GUI tests.

To remember:

• Don't forget unit tests; your code should be covered by them
• Focus more on integration and service tests, where the logic/core features of the SUT can be evaluated
• GUI tests can be simpler to make, yet easier to break and harder to maintain; however, they're also needed 

Manage changes on requirements

Changes on requirements/stories should be reviewed by testers.

• Acceptance criteria may need to be also updated accordingly
• Related Tests may need to be reviewed
  • You may transition the Tests workflow status to some specific status to signal that; a post-function on a workflow transition of the requirement can help with that

Shift Left Testing

Review the requirements with the "testing team", write Tests and run them as soon as possible. By making testing part of the whole development life-cyle, requirements will be more clear, understandable and testable. 

Perform exploratory testing

Define test objectives, goals, and areas of opportunity that allow the tester to use her own creativity and skill.

You can use the default "Generic" test type for this, or create a specific type (e.g., "exploratory") of test kind "unstructured". 

If you want, you may use Xray Exploratory App to assist you on your exploratory testing sessions; more info here.

Requirements/stories should be clear and have acceptance criteria

Clear requirements are essential to have the correct understanding of what aims to be addressed by the requirement and the related business needs. 

Make sure they are reviewed; you can use Jira's workflow status in the related issue for this.

Store Tests alongside your other project development related issues

Although Xray is quite flexible and supports many different project organization scenarios, we recommend you to have your testing related artifacts together with other project related ones, such as requirements/user stories, bugs, tasks, etc. This approach provides a self-contained project, it's easier to understand and manage and, best of all, promotes team collaboration.

Organize Tests properly from the start

Tests must be organized in proper ways, either by using lists (i.e. Test Sets) or folders (i.e. within the project's Test Repository).

But besides this more structured way, Tests can and should be properly "organized" right form the start, independently of how they are to be grouped later on.

This can be achieved by "tagging" the Tests adequately and by identifying the reason for the Test to exist in the first place;  with this you can:

• Assign them labels, component(s) or other fields
• Link them to requirements/stories; if you don't have requirements, consider creating some so you can track your testing results from the requirements/features perspective
• Use a specific field to classify the type of Test (e.g. integration, UI)
• Use a specific field to classify the nature of the Test (i.e. manual, automated)

It's important that your process clearly states how to perform the earlier points, namely the tagging. Otherwise you will end up with similar Tests but tagged in different ways.

Specification

Tests with a purpose

The reason for a Test to exist in the first place should be its purpose/goal.

• Make use of the Summary and Description fields to define the purpose of the Test

All the necessary and clear steps

Clear, non-ambiguous steps (actions and expected results) avoid different interpretations and thus different results.

Tests should have the necessary steps to validate their purpose:

• If you have very few steps, possibly you're assuming too much and you need to detail some further "implicit"/missing steps
• If you have many steps, probably you're validating too many things at the same time; it's best to have tailored test cases for each scenario that you wish to validate; try to have separation of concerns

Mark the Tests so they can easily be found/managed

Use labels or specific custom fields; if using labels, keep in mind that Jira's default label type of fields is case-sensitive and does not provide a way to limit its values; a more restricted custom field type may be more appropriate.

Tests specific for a given Environment

If Tests are specific for a given environment, use the Environment field to clearly identify that.

Whenever scheduling these Tests for execution, make sure to use the Test Environments feature by setting the Test Environments field on the Test Execution.

More info on Test Environments here.

Avoid too many dependencies

A Test that depends on other Tests being executed previously, even by a certain order, makes the Test harder to understand and manage.

Try to avoid dependencies in the first place. If really needed, try to isolate them as much as possible on Pre-Conditions.

Note that a Pre-Condition has only an open text field that is used to describe it; it does not provide semantics to explicit mentioning dependencies to other Test cases.

Missing Precondition/requisite for running a Test

If a certain condition is necessary to run a Test, it's best to abstract in a Pre-Condition.

This makes sense only if you foresee it as something reusable, that is useful for other Tests.

Relative importance of Tests for a given requirement 

Use Priority field to distinguish between different Test cases; Priority is standard field from Jira and thus should be the preferred field for this purpose.

Avoid UI/visual dependencies

Don't make Tests dependent on very specific UI aspects, such as the position or on text labels, unless they're UI/visually aimed.

Test in different configuration scenarios

Perform testing using the same test procedure but in different conditions, such as different combinations of enabled features.

In this case, the preferred approach would be the following one:

• create multiple clones of the same Test
• assign each Test to a specific Pre-Condition, where each one corresponds to a different combination of enabled features

An alternate approach would be using Test Environments and define each one as a specific feature combination. However, this does not scale well if you have several features which would lead to the creation of many Test Environments. Also, you may prefer to use Test Environments as a means to identify the target environment instead (i.e. browser, mobile device).

Few Tests (or few testing) per requirement

A requirement covered by few test cases may be a symptom of lacking test cases, testing only the successful path.

• Create more Tests using white and black box techniques

Dozens of Tests per requirement

A requirement covered by large dozens of test cases may be a symptom that the requirement is too complex or vague.

• Decompose the requirement in more simple, more testable requirements

Avoid having very few Tests, that only validate the obvious

Sometimes testers look at a requirement and write one or two Tests that mimic exactly the description of the requirement, which can be quite simplistic and give a totally wrong sense of confidence

• Have in mind the acceptance criteria and write several test cases for each of them
  • Use techniques such as Boundary Value Analysis (BVA), Equivalence Class Partitioning, Decision Table based testing among other ones, to improve coverage without requiring tons of testing effort

Exercise Equivalence Class Partitioning and BVA

As a means to provide greater coverage, Equivalence Class Partitioning and Boundary Value Analysis can provide enhanced coverage without growing the Test suite too much.

• Clone Tests and use the "data" column to put the values for the inputs of each partition/class
• If doing Equivalence Class Partitioning, don't specify exact values for the class, to have some randomness; if later on you find a bug, you may create a Test case with the exact value to verify the fix

Promote reusability and avoid cloning Tests specifications

If Tests are exactly the same, in theory there should be no need to have clones.

Whenever talking about clones, we may be talking about explicit clones (i.e. cloning the whole Test issue, with all its steps) or implicit/embedded clones (i.e. where the steps of an existing Test are used in another more high-level Test case). 

If you are thinking on using Tests as steps in other more high-level Test case you should consider modular tests and not using implicit/embedded clones.

Depending on the purpose you want to achieve you should use the best suited approach, so if you wish to:

• Perform the same Test but against different data
  • Use test parameterization and datasets, this will allow the correct replacement of the parameters by data defined by the dataset. As you can see in the documentation you can define the dataset in different entities or scopes of your tests.
• Cover/test different requirements, of different projects; in this case projects may be somehow similar and may represent slight variations/conditions of the "main requirement"
  • You may use the same Test for this (probably the Test would reside in some common project); however, if you do so, then the Test result will affect related requirements in the same way (unless you change the Separation of Concerns setting). Instead, and because you're probably really talking about different requirements, the best is to have specifically designed Tests for each individual requirement
• E2E Test composed of steps, where these steps are essentially existing Test cases
  • In this case, if you have defined some Tests that can be used as modules in other Test steps, you should use modular tests to include other Tests as steps of a more high-level Test case. Make sure this will focus the Test in the feature you want to validate and not in the feature of the Tests you are including.
  • You can also design the E2E Test in such a way that its steps are more simple and validate only what is the purpose of the E2E Test itself (e.g. instead of having an E2E Test with all the steps of a login related Test case, the E2E Test can have just one step where the user is expected to login using given username/password)
• Tests having some common initial steps 
  • Abstract those common initial steps in reusable Pre-Conditions linked with all those Tests

Performance considerations

1. Avoid scripted Tests with dozens of steps
2. Avoid Tests linked to many different requirements; try to have just one covered requirement per Test
3. Avoid huge amounts of Tests per requirement; consider more grained requirements

Quick Checklist

Use this to quickly evaluate if you're on the right track or not. This list is a very condensed sum up of many practices mentioned earlier.