This seems like a particularly dangerous linear extrapolation in a nonlinear world. Shouldn’t you all have just had a minimum engagement time that you wanted customers to hit? Like ten hours a month should make you feel comfortable enough that the customer will not be canceling anytime soon. But 200 hours a month might mean you’re causing problems to their lives that they will address by canceling. Even the auto play feature just seem like a standard hack that Goodhart’s Law would predict. User hours may be a good metric to indicate user retention when it’s measuring people’s voluntary use of the product, but could become inaccurate when it measures their psychologically manipulated use.