A CV screening rubric that two consultants score the same way has three parts: must-haves scored pass/fail, nice-to-haves scored on a short weighted scale, and deal-breakers defined by observable evidence rather than opinion. You then prove it works by having both consultants score the same 10 CVs blind and fixing every criterion where their scores diverge.

Without that test, you don't have a rubric. You have two people's instincts written in a shared document.

Why two consultants score the same CV differently

Picture a client brief that says "strong stakeholder management, ideally from a regulated sector." One consultant reads that as "has managed senior clients." The other reads it as "has worked in banking or pharma." Both score the same CV, one gives it a 4 and the other a 2, and the shortlist depends on whose desk it landed on.

Nobody is being careless. The brief used words that mean different things to different people, and fatigue makes it worse. By CV 120 on a Monday, you're scoring against a vague memory of the criteria rather than the criteria themselves. The failure mode is quiet: the client sees a shortlist that looks fine, but two equally qualified candidates were ranked differently for no defensible reason.

Build the scorecard from the client brief

Start with the brief, not a template. A job description template helps you capture what the client actually asked for, but the rubric has to be specific to this mandate.

  1. Extract must-haves. These are the three to five things without which the candidate can't do the job on day one: a licence, a minimum level of a specific skill, a right to work. Score them pass or fail. If you can't write the evidence you'd look for on a CV ("has run a team of 5+ for at least two years"), it isn't a must-have yet.
  2. List nice-to-haves with weights. Keep it to five or six. Give each a weight of 1 to 3 and score the candidate 0, 1 or 2 against a one-line definition of each score. A 2 means "named tool used in a role for 12+ months," a 1 means "mentioned, with no duration or context," a 0 means absent.
  3. Define deal-breakers narrowly. "Job-hopper" is an opinion. "Four roles under 12 months in the last six years, no contract or maternity explanation" is a deal-breaker two people can check. The same goes for salary expectations above the client's ceiling or a notice period beyond what the client can wait.
  4. Write the evidence line. For every criterion, note what counts and what doesn't. This is the part most rubrics skip, and it's where agreement comes from.

Test it: 10 CVs, two consultants, blind

Pull 10 CVs from a previous mandate or the current pile. Include a few obvious yeses, a few obvious nos and at least three borderline ones. Each consultant scores all 10 independently, with no conversation and no peeking. At 90 seconds a CV, that's 15 minutes each, 30 minutes of combined desk time to find out whether the rubric holds before you commit 200 CVs to it.

Measure the disagreement

Put both score sets side by side and check three things:

Then discuss only the flagged criteria. Ask each consultant what evidence they used. You'll nearly always find the same thing: the criterion was an adjective ("strong," "relevant," "senior") rather than a observable fact. Replace it with something checkable, add an example of what a 2 looks like, and re-score only the CVs that caused the gap.

The rubric is also your AI brief

A rubric this specific is the same document an AI screening tool needs. Vague briefs produce vague matches, whether the reader is a consultant or software. If you paste your must-haves, weighted nice-to-haves and deal-breakers into CV Matcher, it scores against the same evidence lines your consultants use, which means you can compare its output to your blind test and see where it disagrees with your team. Our guide on structuring your job brief for AI screening covers the formatting in more detail.

Where the human judgment stays

A rubric standardises the first pass. It can't tell you why a candidate left three roles in four years, whether their salary expectation is flexible, or how they'll land with this particular hiring manager. That's what the qualification call is for, and it's worth reading about what "qualified" should mean after an AI screen before you decide which candidates get a call.

Your first step on Monday

Take the next brief that lands and write the scorecard before you open a single CV. Run the 10-CV blind test with a colleague, fix the two or three criteria that split you, then screen the full pile. If you want to see how the rubric performs at volume, Start Free Trial and run it against a live mandate.