I think we could call this black-box technique an advanced one developed from pairwise testing. Unlike pairwise which just offers us to test pair of all factors, we can use classification tree method to increase coverage by testing triples, quadruples… of some factors. In short, this technique and tool allows us to design test cases more flexibly if based on our risk scale, we want to test some factors more heavily than others.
Supported tool:
- CTE XL – Download here (you need to fill in some of your name, email, get license by link and email before using it)
Example:
A a web application based on 3 factors as below:
• The first factor: connection speed (2 options: dial-up and broadband)
• The second factor: operating system (2 options: Windows XP, and Windows Vista.)
• The third factor: browser (2 options: Firefox, Internet Explorer)
a) Design test cases for all pairs possible
b) Design test cases for all pairs of OS and Browser only
c) Design test cases for all triples possible
Results:
Note: For scenario a, what we got is exactly when we use allpairs tool. Scenario b, it’s because we just focus on pair of OS and Browser only, all pairs possible of these 2 factors are ensured while other factor (Speed) is ignored. Scenario c, using threewise method to generate all triples possible, we ensure all combination of every 3 factors existed in our test suit. Here, because we only have 3 factors with 2 options each, number of test cases is the same when we use completed combination which is calculated by 23= 8.
Posted by Khuong Nguyen 
