Tuesday, July 05, 2005

Should All Requirements Be Testable?

When product managers specify requirements for a product, we strive to formulate the requirements in such a manner that they are testable. Companies obviously want to know when they have developed a product that satisfies the requirements; to this end we want to provide testable requirements. Yet sometimes it is not practical to test a requirement directly.

Take, for example, the requirements for a new, super-reliable model of car. We might specify that the car should last seven years without repairs as long as the owner maintains the car according to a certain maintenance schedule and doesn't have a collision. But it is not possible to directly test whether the product meets these requirements without producing the car and driving it for seven years.

The difference is between requirements that are possible, in principle, to test, and those that are possible, in practice, to test. As product managers, we should strive for requirements that are possible to test in principle. We can't ignore market demands just because it's hard to test whether a product satisfies them.


Marcus Ting-A-Kee said...

For untestable requirements, you could similate approprimate tests. In your car example, this may mean stress testing components, similating weather conditions (e.g., putting the car in sub-zero conditions), starting it up 8,000 times and driving 60 miles each time etc... While not an exact match it might be close enough to give you some certainty.

Roger L. Cauvin said...

I agree completely, Marcus.

Just because it is not practical to test a requirement directly doesn't mean you can't test it indirectly through simulation.

A requirement always prescribes a test. Sometimes this test is not practical to carry out before releasing your product. In such cases, your team devises and carries out other, related tests that determine indirectly whether the product satisfies the requirement.