This is a really good point Jim. I would classify what did not work as “good data” as long it is accurate. I think the problem with using what did not work is that it is hard to determine what caused things to “not work.” It could be one thing that caused failure — and everything else could be “right.” I value when people keep trying to succeed, through failure after failure, and then actually succeed. Those people can give a fascinating glimpse into what was changed in order to succeed.