Cost per hire vs cost per screen — the two numbers HR confuses

These two numbers used to move together, which is why people still treat them as one. They have come apart. Cost per screen has fallen to a few paisa; cost per hire has barely moved, because it was never mostly about screening. Budgets built on the old relationship now produce the wrong answer in a predictable direction.

The two definitions

The first has a denominator of applicants. The second has a denominator of hires. They differ by however many people you looked at per person you employed, and that ratio is not stable — it moves with the posting, the channel, and the season.

What each one contains

Cost per screen is the narrow one: reading the document, evaluating it against requirements, recording the result. When a person does it, that is three minutes of attention. When software does it, it is a few paisa — our own ledger puts the median at $0.0028 per CV scored across 65 runs, and the published client rate is ৳2.50.

Cost per hire contains that, and then everything else: the job posting, every hour of every interview, the offer process, and the hours spent deciding. Broken down line by line, the screening portion is the smallest item on the list and the team's time is the largest.

The ratio, and why we will not give you a benchmark for it

In our own data, one hire consumed 26 screened CVs — 78 real candidates, three of them marked as hired.

Three hires is not a benchmark and we are not offering it as one. It is one company, one month, three events; the number would move wildly with a fourth. We are showing it only because the shape of the arithmetic matters more than the figure: whatever your ratio is, cost per hire is cost per screen multiplied by it, plus everything screening never touched.

The trap this sets

Here is the mistake, in the order it usually happens:

1. Screening gets cheap, so the cost of an extra applicant looks like nothing.
2. The posting gets widened — more channels, softer requirements — because why not.
3. Applicants double. Cost per screen stays near zero, so the dashboard looks fine.
4. Cost per hire goes up, because the shortlist got longer, the interviews multiplied, and the deciding took twice as long.

Every step is locally rational and the result is worse. The number that was supposed to warn you — cost per screen — is precisely the number that cannot see the problem, because the expensive part happens after screening ends.

This is why being able to process 500 applications quickly is not the same as being able to hire from 500 applications. The processing is half an hour. The deciding is not.

What to measure instead

• Cost per hire, with your own hours in it. The only number that answers "should we do more of this". It needs your hourly figure, which only you have.
• Interviews per hire. The most sensitive early signal. If it rises, your shortlist quality fell — usually because the requirements got softer.
• Time to first conversation. Not time to hire, which is dominated by notice periods you do not control. The gap between someone applying and someone speaking to them is yours, and it is where good candidates are lost.
• Cost per bad hire, separately. Nobody can give you a transferable figure, but your own records can give you a range, and that range is what tells you how much evidence is worth gathering.

What cost per screen is still good for

One thing, and it is a real thing: deciding whether to look at everybody.

When screening cost three minutes of attention each, "read the first sixty and stop" was a defensible policy. At a few paisa it is not — the cost of looking at applicant 400 is now genuinely negligible, so the only reason not to is that nobody built a way to. That is a change in what is reasonable, and it is the one decision this metric should drive.

What it should never drive is a target. "Reduce cost per screen by 30%" is a goal you can hit by changing nothing that matters, and it is entirely compatible with hiring getting more expensive.