> In predictive analytics and machine learning, concept drift means that the statistical properties of the target variable, which the model is trying to predict, change over time in unforeseen ways. This causes problems because the predictions become less accurate as time passes.
This is true, but...meh? OK? Perhaps there are places where inefficiencies are not a bad thing, and perhaps advertising is one of them.
Perhaps, even, we should require people to think and empathize and communicate rather than feed a pile of unvariegated data into a black box and do what it tells us to.
I'm shockingly OK with having to make a connection with people to try to sell them something, and to have that be an upper limit on a company's reach.
IF a model is trained with historical data, that would be good enough to track without actually tracking.