How to Write a Candidate Scoring Rubric Your Recruiters Actually Trust
· 8 min read
Turn "I know a good candidate when I see one" into a scoring rubric an AI screener can run — without losing the judgment that makes your desk money.
Every good recruiter can look at a CV and know, in seconds, whether it is worth a call. The problem is that this knowledge lives in their head. To let an AI screener do the first pass, you have to get it out of their head and onto paper — as a scoring rubric. Here is how to build one your recruiters will actually trust, without flattening the judgment that makes your desk money.
Start with the two kinds of criteria
Every rubric is made of two things, and mixing them up is the most common mistake.
- Knockouts are pass/fail. No valid work visa, no required licence, wrong country — these are not "worth fewer points," they are disqualifying. A candidate who fails a knockout is out, full stop.
- Scored criteria are matters of degree. Years of relevant experience, depth in a specific tool, industry background. These get a weight, and candidates earn a partial score against each.
Keep them separate. If you turn a knockout into "worth 40 points," you will eventually rank an unlicensed candidate above a licensed one, and your recruiters will stop trusting the list.
Weight what actually predicts a placement
For the scored criteria, assign each a weight, and make the weights add up to 100%. The discipline of forcing them to total 100 is what turns a wish list into a rubric.
Ask a sharper question than "what do we want?" Ask: what has actually predicted a good placement for this role in the past? Often the honest answer is two or three things carrying most of the weight, and a long tail of "nice to haves" that barely move the decision. Weight accordingly. A rubric where everything is important is a rubric where nothing is.
Write down the unwritten rules
This is where the real value is. Sit a recruiter down and ask what makes them pass on a CV that looks fine on paper. You will hear things like:
- "Three short stints in a row, for this kind of role, is a flag."
- "If they have only ever worked agency-side, they usually struggle with this client."
- "No mention of the certification means they probably do not have it, even if the experience is there."
These are the rules that separate your screening from a keyword search. Capture them. Most never get written down because nobody ever had to — the recruiter just knew. The rubric is where they finally get written.
Watch for criteria that are really proxies
One caution. Some criteria feel job-related but are actually standing in for something you should not screen on — a proxy for age, background, or something unrelated to doing the work. "Graduated in the last five years" is not experience; it is age with extra steps. Flag anything that is not clearly about the job, and either rewrite it to measure the real thing or drop it. A good rubric is fairer than gut feel because it is written down and can be checked.
Test it before you trust it
A rubric is a draft until it survives contact with real CVs. Run it against a batch you have already screened by hand and compare:
- Did the candidates you would have called rank near the top?
- Did anyone you would have rejected sneak up the list? If so, which criterion let them?
- Did a strong candidate score low? That usually means a weight is off, not that the candidate is bad.
Adjust the weights, tighten the knockouts, and run it again. Two or three rounds is normal. This is exactly the tuning step we run in the first weeks of a build — the rubric is never "done" on day one, and it should not be.
You do not need documentation to start
If your reaction to all of this is "we have never written any of it down" — good, you are normal. Most firms have not. You do not need documents; you need to be able to say what a good candidate for a role looks like. Turning that into a rubric an AI can run is the job of the free audit. We pull it out of your recruiters' heads so you do not have to.
For the bigger picture of where the rubric fits, see how AI CV screening cuts shortlisting time, or why your ATS's built-in AI still misses your best candidates.