We launched a 1:1 tool for sales managers that used predictive analytics to suggest which deals needed to be discussed (ie Deal X looks like it's stuck). The reason being, when I was a sales manager, I found people would naturally talk about positive stuff rather than easily giving the whole picture.

It failed because quite simply, when we tried to get other sales managers to use it for free (hoping it'd be popular), we found they wouldn't make the time to spend 2 mins planning a 1:1, probably as they're managed on 'selling deals, right now' in many cases, which is a bit sad!

Learning: I should have read the Mom Test.

With hindsight, I wish I'd built it for engineers based on Pull Requests (PR #181 had a ton of back and forth / NLP shows it got heated). It would have been fun if nothing else!