A few months back a client emailed asking how to do a predictive thing - he was trying to build a model to predict sales so that we could find out how much impact non-modelled factors had on sales, like outside advertising or real-world events.

It was explained. He said it was too simple, and he'd do it himself. We didn't object.

He emailed again a few days ago to explain that he wanted us to look through it.

He'd tried to use revenue as a predictor for sales.