>But it's still a human making a decision in the end
I've got a story here from a previous job that illustrates this.
I used to be part of an Analytics and Forecasting team at a well known non-tech company. It's a global company that sells a lot of things to consumers. This means they strive to have a very good understanding of sales and manufacturing volume because that drives everything from raw material orders and supplier contracts to how much of product to put where so people can go and buy it.
We had short term forecasting teams generating forecasts (at a brand level) and analysts (mostly aggregations, not really statistical inference) that rolled into a medium term analytics/forecasting team(s), that rolled into a long term analytics/forecasting team(s). Not to mention Research teams that did pure statistical and economic modeling to try and understand market factors in a more comprehensive way so they could better inform all these forecasts. All of this went right to the top of the food chain.
The glue between all these teams were managers. And one of the fundamental problem for managers with all of this comes from this simple question - if one of these teams provides a forecast that is very different from the rest, what does that mean? Is that team right or wrong? And how does that information now flow into the other teams so they can incorporate it into the info that goes upstairs.
From this question emerges an astonishing amount of group think. Not just from the company I worked at, but at all competitors as well.
When companies ask "Do we forecast sales to go up" they are really asking "Do we think this market segment is going to go up in the future and we have the right strategy to move up with it?" Cars, Snacks, phones, whatever the market segment, you're really asking how you are doing compared to your competitors in the space.
To understand where you are going, you need to understand how your competitors are doing. This data comes from agencies and consulting houses. But since everybody uses the same data from the same agencies, everybody is feeding each other. What can emerge from this is a massive amount of group think.
I'm greatly simplifying everything of course. When these teams get it right they do amazing stuff like making sure the thing you want to buy is available when you want to buy it. But when they get it wrong, it seems like most everybody else gets it wrong as well. This can slowdown an entire industry as everybody realizes they got it wrong.