Selection Accuracy and Base Rates

A hundred applicants. Twenty of them would succeed in the job. Your process catches 80% of the good ones and rejects 80% of the poor ones — which by any normal description is a good process.

It passes 16 good people and 16 poor ones. Of the 32 you hire, half work out.

That is not a criticism of the process and it is not a rhetorical trick. It is exact arithmetic, and it happens because there are four times as many poor candidates as good ones, so even a small error rate on the large group produces as many false passes as the small group produces real ones. It is the same reason a test for a rare condition mostly returns false alarms.

The page also counts what almost nobody counts: the four good candidates it rejected. Twenty per cent of everyone who could have done the job, gone elsewhere, and you will never find out. That is the price of a stricter method and somebody should be looking at it.

Then it shows the finding that changes what you would do. The same method, on a pool where half the applicants would succeed, hires people who are 80% right — and a much sharper method on the original pool does worse than that. Where the applicants come from is doing more work than the sifting is, and it is not where the effort goes.

It treats accuracy as the two numbers it really is: how often the method passes a good candidate, and how often it rejects a poor one. A single accuracy figure quietly means whichever of the two flatters the method, and both boxes are on the page.

It makes no claim that any method has any particular accuracy. You type what you believe about your own process, and the page shows what that belief implies — which is often the first time anybody has looked at it.

Save as many versions as you like and put any two side by side, figure against figure, with the better one on each line marked and the difference in its own column. That screen is the one that decides things, and it is the part a free calculator cannot do at all. The same method on two different pools is the comparison this exists for.

One HTML file. No account, no subscription, no internet needed. Opens on Windows, Mac and Android; the readme explains the extra step on iPhone and iPad, where Apple blocks JavaScript in downloaded files and every tool of this kind is affected.

Everything lives in your browser on your device. Nothing is uploaded, which is why it works offline and why nobody else can see your figures — and it also means clearing your browsing data deletes it. The tool does not pretend otherwise: the header shows how long since your last export at all times, turns amber after three days and red after seven, and export is one click that produces a plain file you keep.