> So if the developers knew to traverse a binary tree, how many of the 75% succeed? I would say it's necessary but not sufficient. There is no perfect intervie

> So if the developers knew to traverse a binary tree, how many of the 75% succeed? I would say it's necessary but not sufficient. There is no perfect interview strategy. But for projects that are entirely developed by people who can't traverse a binary tree, I'd say the odds of failure are very, very high. Sure, a strong developer can hit that 80% of performance quickly (in probably less than 20% of the time), but a weak developer who doesn't know how to optimize probably won't even make 5% of "maximum potential performance." I had to work with a tool once that had a "process" function that would take 2-3 minutes to do its work. It was used by level designers, and it was seriously impeding their workflow. The tool was developed by an game developer with something like 10 years of professional development experience (he was my manager at the time), and he thought it was as fast as it could go -- that he'd reached that maximum potential performance, with maybe a few percentage points here or there, but not worth the effort to improve. He didn't want to spend another 80% of his time trying to optimize it for a minor improvement either, so he left it alone. I looked at what it was doing, figured out how it was inefficient, and in less than an hour rewrote a couple of key parts, adding a new algorithm to speed things up. Bang, it went from minutes to 200ms. No, I'm not exaggerating. Yes, I had a CS background, and no, the experienced developer didn't. If you end up with accidental n^3 algorithms baked into your architecture, 5% performance in development can be 0.1% performance in the real world, or worse as your N gets large enough. And yes, that's even if you index your data correctly in your database. And that's when your site falls down the moment you have any load, and you end up trying to patch things without understanding what you're doing wrong. In my example above I improved the speed by nearly a factor of a thousand. In an hour. That can easily mean the difference between holding up under load and falling over entirely, or the difference between being profitable (running 4 servers that can handle all of your traffic) and hemorrhaging money (running 4000 servers and the infrastructure to scale dynamically). Which is why you need a strong developer to run the project to begin with. Maybe you're that strong developer; I'm not really speaking to you in particular. But I know a lot of developers who just don't have the right background to be put in charge of any projects and expect that they'll succeed.