I agree that completed features is ultimately a far better measure of progress than just tests.

Although my point was only that it makes *more* sense to track passing tests than expended effort on 'tasks'.

What I had in my mind was the typical measure of progress: tasks on a task-board and effort expended on a burn-up/down chart… I think a step in the right direction would be to place less emphasis on those things and measure something that gives us a better indication of progress. I think passing tests are a better indication of progress than tasks. Just like miles travelled/remaining on a journey is a better indication of progress than the amount or fuel-burned/remaining. Useful as that latter is, it isn't an indication of progress.

I wasn't saying it was the best and only way of measuring progress

I've not worked on any projects where automated checks are implemented entirely in advance of programming the feature… if people are doing that, they're not doing xDD (TDD/ATDD/BDD). These are iterative and incremental learning processes and, in my experience, do not work well if you try to specify all the checks in advance.

Thanks again for the comment.

It doesn't make sense to measure progress with passing tests, whether you call them checks or not. It might make sense to do that if the development of a software product were linear, or a set of pegs that you insert into a board. But it isn't.

A passing test (and in particular, a passing check) tells you that a product is capable of producing a correct result in a specific, highly controlled set of conditions. It tells you that the product can do something. It doesn't tell you, or even warn you, of terrible problems in the product of which the check is unaware. For that, you need exploration. As J.B. Rainsberger said once, a green bar doesn't tell you you're done; it tells you that you're ready for a real tester to kick the snot out of it.

It does make sense, though, to measure progress in terms of completed features, which I think is what you're saying, where “completed” means that the feature has been developed and checked and tested to the degree that a person has deemed it acceptable. I see great risk if the person has delegated that decision to checks alone, especially when the checks have been devised entirely in advance of the programming work on the feature.

—Michael B.