I suspect something like this happened: - The team behind the zestimate kept touting how accurate it is. Since they needed to keep improving it to keep their j

I suspect something like this happened: - The team behind the zestimate kept touting how accurate it is. Since they needed to keep improving it to keep their jobs they kept cherry-picking the best results to showcase. Edge cases were silently not mentioned and failures were spoken of as little as possible. The further up the management chain the result went the nicer they looked as each layer tweaked them to be even more impressive sounding. None of it mattered much so who cares about a few white lies? - The other executives seeing these great results decided to create a new business model based on top of it. The DS team and their management was not consulted about this until many wheels were already moving. - The zestimate team and their management chain would get fired (or otherwise punished) if they suddenly showed all the edge cases and potential failures they never mentioned before. So they collectively kept quiet, hoped they could fix the edge cases and hoped it would all go well. Messengers tended to get shot in this situation and no one wanted be that messenger. edit: In my experience skewed institutional incentives are more often the source of issues than technical mistakes or low level employees being blind to things. You don't get promoted in most companies by telling the CEO they're wrong so people don't.