scaling backend servers: what nobody explains until you're already on fire

Most content about scaling backend systems describes what to do after you've already identified your bottleneck. "Just throw more servers at it." "Use a message queue." "Cache aggressively." All fine advice. But nobody explains the step before that — how do you actually figure out what your bottleneck is? And once you know, which of the dozen scaling techniques do you reach for? I've been through a few of these situations now and the same pattern keeps showing up: the problem is almost never wha