Basically, there's a ton of completely unqualified (or even under-qualified) applicants who apply to just about every job opening out there. If you've ever been on the other end of the table, posting a job opening and sifting through applicants, the situation becomes clear very quickly.
I'm actually surprised 250 make it past their initial screen. The rest of the numbers seem reasonable to me though.
A few companies do manage to do it, but as far as I can tell you have to set up your corporate culture, HR department etc. from the outset to be prepared and ready for firing people frequently, and that's not something many startups (other than maybe Netflix?) were/are really well prepared for...
Firing doesn't seem hard but I think your point does stand that the cost of having to do so is kind of great. Find, hire, onboard and train a new employee, one that you hope does better than the bad hire.
We have benchmarks to understand how well a company is converting at each step in their funnel. CircleCI is converting well in 2 of their funnels, but the other 2 are underperforming significantly per my other comment in this thread.
Past a certain level of skill, it really does come down to specific experience, priorities, communication styles, etc. There are many worthwhile devs that simply have a less ideal package for that specific role, at that specific company, at that specific time.
It still seems high, but I can imagine these average numbers being inflated by some important 'Lead Engineer' type of position, for which the extreme vetting is reasonable.
> Our ultimate goal is to have the best people here, doing their best work.
I feel like everyone is so afraid of hiring a 0x or -1x programmer that they don’t want to take a chance on a mere 1x programmer.
If I ran a company (I don’t!) on the scale of hundreds or thousands of developers I’d be very choosy about who I let into senior/leadership type roles, but be rather accepting of individual contributors, the hope being that many of them would grow and develop their skills.
Am I just vastly underestimating the overal strength of a random sample of 1000 candidates? Are there really only 3 worthwhile developers?