Friday, December 23, 2005

Risk based Testing ....some points to ponder

Risk based Testing - I can see lots of eyes rolling and eye brows raising on seeing this term. Likes of Project managers and Delivery managers simply love to hate this term "Risk". My boss, Manjunath Hebbar, other day gave a this beautiful definition of risk "A risk is piece of information regarding tomorrow 's (future) problem as on today (present)" and he further said - in contrast this an issues is "problem in the hand". Well - this discussion with him was around "project management" not about testing.I am not that competent to talk about project management, so let me be happy with what I am good at - testing.

Risk based testing is all about "Test" the risks in the object that you are testing. That is simpler while saying. In reality - Risk based testing involves

1. A methodical approach for identifying risks
2. Design the tests to explore the risks - know more about identified risks - As quickly as possible
3. Precipitate the failures resulting from the risks - execute tests - As quickly as possible
4. Explore other possible (unknown in the beginning) risks - As quickly as possible
5. Repeat steps 1..4 until all risks evaporate to the reasonable degree OR you run out of time/budget.

If you claim to have mastered Risk based testing technique - check you have following -

1. A proven method of identifying the risks in the product domain - most of these proven ones have a strong theoretical or empirical back ground ( mathematical/statistical or similar). A method you can rely on.
2. Test design methodology - that is matured enough to design tests to explore and precipitate this risks
3. An exit Criteria to know when are "reasonably done with"
4. Knowledge of "Risks" associated with this Risk based ( ironical isn't?) approach to testing.

There seem to be two broad categories of Risk based testing methodologies

1. Exhaustive Risk modeling and analysis - those with involved mathematical and statistically models one the lines of those used for "Reliability Engineering"

2. Heuristic Risk based Testing - rather simplistic one. James Bach has written an excellent article on this topic. As described in the article this is one of the approaches that can be adopted ( in combination with other "formal" techniques) when you have resource and time crunch. He cautions that the whole concept of "Heuristic" could be a fallible one and more of provisional and plausible. It tends to explore the solution.

I happen to discuss this issue with others in Google group on software testing, here. There are some very interesting and thought provoking ideas from Andy Tinkham in this thread.

Read following lines carefully - before you finish reading this post and tell me what do you think …
1. Testing is more about finding important problems that threaten the value of the product - issues that bug someone who matters in the context
2. Risk is a problem that might happen - higher the probability of Risk - higher is the probability of failure.
3. Should not all testing be risk based? Why would one want to test where there is no risk? When non risk based testing would be appropriate ….



Anonymous said...

With no condescension intended in this comment, I suggest that you should consider having someone help edit your blog. The information seems apparent but is malformed due to ineffective communication.

Shrini Kulkarni said...

Dear Anonymous,

thanks for the comment. I really appreciate your critical view of my blog content...

Please post your mail address ..

Let us have 1:1 discussion - if you wish to share your views and help me in improving my blog content ...

I suggest you that when posting such comments - please state your name at least ....



Anonymous said...

hi there,
Its true that all testing is risk based, any one of the following would be on track, like time, cost or resource. So all are risk based testing. But Actual Risk Based testing to find out the most vital or critical part of the project and then work on that for testing, but its not sure that other part would not result in critical error. So its all about probability of what we test and how. :)