Here's a little thought experiment: Let's say the very most talented potential hires almost never apply for a job, because they get headhunted. That means the p

Here's a little thought experiment: Let's say the very most talented potential hires almost never apply for a job, because they get headhunted. That means the pool of applicants for an open position will mostly be a mix of competent-but-not-brilliant (or at least not-yet-proven-brilliant) and incompetent people. The hiring process, if it (mostly) works will lead to hiring the competent people while leaving the incompetents behind, who are then free to apply for job after job until the hiring process somewhere fails to filter them out. The point is, the incompetent people will spend much more time applying for jobs than the competent people will. This also means that at any given time, the incompetents will make up the vast majority of applicants to any given job - in other words they are vastly overrepresented in the 'market supply' that a company gets to see. This leads to some interesting effects. One is that many companies claim to hire only the top 5%, when in reality it just means that they hire only the few competent people in the pool that they see (and these competent people are more likely to be e.g. the top 50%-5%, since the actual top 5% got headhunted). Another is that if a company used a naive metric of the 'market supply' to determine things like the demographics of the available workforce, that metric would mostly be comprised of the demographics of said incompetent people. Which might lead to some interesting biases (IIRC it's fairly well documented that men tend to overestimate their abilities quite a bit more than women do - the saying is, if a position requires 5 skills, a man will apply if he has 1 of them, and a woman will not apply if she has only 4 of them). The point is, if you wanted to do something like benchmark a company's internal diversity against the market supply, you might need a better way to estimate that market than just looking at applicants, since that is mostly noise and very little signal (and the point of most hiring processes is to filter out all that noise)