Updated: Jan 26, 2022
The graphical user interface (GUI) is the primary platform for manipulating software. Due to its importance to the end-user, it must be carefully tested to ensure that the user can easily interact with it, perform highly complex operations as efficiently as possible and ensure that it is clear enough for them not to ask technical questions.
When testing a user interface, the team should follow guidelines that include the following items:
Ensure that the application logic makes sense from a customer point of view.
Simplify the use of complex functionalities.
If the user needs to learn the software’s graphical interface from a thousand user guide pages, then this GUI is not a usable interface.
Tests must represent customer expectations.
The GUI must have the right look and feel.
The user interface should do what it is supposed to do and nothing else.
Use positive and negative tests to ensure that objects can handle different inputs.
How much testing is enough?
It is easy for testers to get carried away and waste critical testing time spent in other areas of the application. How much testing is enough, considering the importance of the user interface? The simple answer is that just like any other test type, stop testing when you have run all the tests to mitigate the risks identified before testing.
Should GUI testing be automated or not?
To answer this question, we need to consider a few issues first:
Even the simplest user interface can become complex to test, as it contains almost infinite supported scenarios that are a candidate for testing.
GUI testing is one of the most time-consuming tasks in the software industry.
Automated frameworks will never be able to cover test scenarios related to the look and feel of the application.
The GUI can be changed multiple times through the development process, especially in an Agile environment. As a result, the maintenance cost of UI automation is often not worth the time.
There must be a very good justification for automated UI tests. In my opinion, there should never be exhaustive automated coverage for this type of test, as that would add less value than automating other components of the application.