ML is best suited for situations where there is no practical solution using typical statistics techniques and where marginal improvements in accuracy lead to si

ML is best suited for situations where there is no practical solution using typical statistics techniques and where marginal improvements in accuracy lead to significant boosts in revenue or some other useful metric. It turns out there aren't that many of those problems unless you're operating on truly enormous scales.