Comment by sph

> I would say that a reasonable person could have foreseen Concord failing.

> Then you add on that they already missed the train for hero shooters by about eight years

There's an analysis video I saw on Youtube which touches upon this fact: data-driven production suffer from two major problems: when they register a signal (i.e. people like hero shooters), it's already too late. It is impossible to catch a growing trend, just one that has already reached its peak.

The second problem is that the data is misleading in the first place. Using social media sentiment, for example, is pure nonsense because social media personas are not real people. They don't buy toilet paper, they don't talk (or Google) about 99% of their boring existence, and the signal is manipulated by bots and malicious agents. It is absolutely crazy that companies still haven't learned that you cannot hire a bunch of data scientists and predict the next major hit. They have tried for the past 20 years, but human ingenuity (and the chaotic human psychology) is totally elusive to their silly models.