Line by line I rewrite the requirements and implementation details in Excel, putting each word or phrase in its own cell. In the row above each word or phrase I put a Category name, and below each word I list an alternative from the category.
Then I read the sentence substituting each word in the list, and noting my expectation of the outcome. This way I map my coverage space and oracle space, and learn what elements make a difference.
After I summarize the spaces, I pick a combinatorial algo to combine inputs and expected results as cheaply as possible. I write a script that expresses the gist of the oracle. This could be an Excel formula, Python script, whatever.
I hand off the test coverage map, oracle algo, and the preliminary list of cases to the dev team so they can use it to guide their functional and testability design and implementation.
Then when I test the system I track the coverage elements touched by the tests, and light up the map. Each test can either extend coverage across the map, or dig down deeper in a spot on the map. Wherever the biggest deal seems to be that’s where I go next.
Once it gets going I hand off to my team to wrap up and I huddle with the business to model the next problem space. Soup to nuts this process can take 2 hours to 2 weeks - or longer - depending on everything.
