How can you reduce risk when implementing agile development in your SDLC?
There’s a nice post from yesterday by Michael Feathers (author of the classic Working Effectively with Legacy Code) about characterization testing - the process of writing code that characterizes the actual behavior of that code. He makes the important point that, “When a system goes into production ... it becomes its own specification.”
In crowd testing, we see this point play out on the macro level. Since our testers bring their own testing heuristics [PDF] with them, they approach software in ways that its developers never envisioned. So sometimes customers will come to us with a production app they think is solid and we discover hundreds of “bugs.”
It then falls to the customer to determine which of these are “bugs” and which aren’t; a decision of no small moment to the testers, who get paid on the basis of the bugs they find. But Feathers’s article makes me realize that in some cases, our testers are effectively discovering the spec for the customer.
They found something the customer didn’t know about, which the moment the “not-a-bug” button is pressed our app becomes forever more a “feature.” You’re welcome.