The LTV point is interesting, and depressingly true. The LTV of a customer is a metric that, before maybe a few decades ago, most companies simply did not calculate. But because today we can use computers to track, correlate, and analyze purchases, we invented the LTV metric, and now it (among others) circles back into how the company treats customers in the first place. If you have a low LTV today, you'll get less support, which means you're unlikely to spend more with them.
Its a pattern depressingly similar to Artificial Intelligence (after all, Elon Musk has made the compelling argument that companies are just AIs of a different form). If you train an AI to predict future possibility of committing a crime based on someone's face, using data of past crimes/arrests/etc, the output will likely, unfortunately, disproportionately rate Black individuals as likely to commit. When deployed improperly, this would lead to increased surveillance and enforcement, more crimes are noticed, more confrontations, accidents happen, and the prediction is suddenly true.
The past does not predict the future; it writes it, and humanity is the pen. The outcomes these predictions promise aren't intrinsically true; they become true because they become recursive inputs into some networked intelligence system, whether that be a computerized artificial intelligence, a company, a government, a market economy, or even humanity as a whole, which then, in adversarial situations, optimizes for self-preservation.