Sourcing

Candidate scoring in 2026: how to score and rank candidates against a role before you contact or interview them

A method to score and rank a long list of profiles against one role: must-have filters, weighted criteria, a scale with anchors, tie-breakers, AI-assisted ranking and the human review that keeps it fair.

·9 min·The HeyTalent Team · Recruiters & Product
Sourcing

Candidate scoring in 2026: how to score and rank candidates against a role before you contact or interview them

Candidate scoring means giving each profile a score on criteria tied to the role, with weights decided before you look at anyone, so you can rank the list and decide who to contact first. It works in four moves: must-have criteria act as a pass/fail filter, desirable criteria get a weight, each criterion is rated on a scale with written anchors, and ties are broken by a rule set in advance. AI can apply those criteria to every profile in a long list, but the recruiter defines them and reviews the result, especially the rejections.

What is candidate scoring

Candidate scoring is the practice of rating each candidate against the requirements of one specific role, criterion by criterion, and adding those ratings into a single score that orders the list. It happens before the interview, during sourcing and screening, and it answers one question: which profiles deserve the recruiter's time first.

Candidate ranking is the output of candidate scoring: the list sorted from the highest score to the lowest. A candidate scorecard is the sheet that holds the criteria, the weights and the scale, and it is reused for every profile in the same search.

Candidate scoring is not interview evaluation. Rating answers in an interview uses other evidence and another kind of sheet, covered in our evaluation checklists for interviews. Pre-interview scoring only uses what a profile or an application shows: experience, skills, location, languages, career path.

How to score candidates step by step

To score candidates, a recruiter writes the criteria from the job brief, splits them into filters and weighted criteria, rates each profile on a fixed scale with anchors, adds the weighted scores and breaks ties with a rule decided in advance. The whole scorecard is written before the first profile is opened, so the list cannot reshape the criteria.

  1. Write the criteria from the brief. Each criterion of a candidate scorecard comes from a requirement of the role agreed with the hiring manager or the client, not from what the first good profile happens to have.
  2. Separate must-haves from desirables. Must-haves are filters; desirables carry a weight.
  3. Set a scale with anchors. Every point of the scale has a written description.
  4. Score every profile with the same sheet. Same criteria, same order, same evidence.
  5. Rank, break ties and review. Sort by total, apply the tie-breaker and have a person review the top and the rejections.

Must-have criteria as a filter

A must-have criterion is a requirement without which the candidate cannot do the job, and in candidate scoring it works as pass/fail, not as points. A senior backend role that requires Go in production, or a sales role that requires fluent German, excludes profiles that miss it, whatever else they score.

Keep the must-have list short: two to four criteria. Every extra must-have shrinks the list and tends to import preferences that are not real requirements, such as a specific university or a former employer.

Weighted criteria and points

Desirable criteria in candidate scoring get a weight that reflects how much each one matters to the role, and the weights add up to 100 % so totals stay comparable across profiles. A common split gives the most weight to the core skill of the role and less to secondary skills, sector experience or location.

Designing the weights is the same exercise as building a selection matrix, explained step by step in our guide to the candidate selection matrix. For candidate scoring, the only extra rule is that weights are fixed before the first profile is scored and do not change during the search.

A scoring scale with written anchors

A scoring scale for candidates works when each value has a written anchor that says what evidence earns it. A 0–3 scale is usually enough for profile screening: 0 means no evidence in the profile, 1 means indirect evidence, 2 means clear evidence, 3 means clear and recent evidence at the level the role needs.

Anchors are what make two recruiters give the same score to the same profile. Without them, a "2" from one person and a "2" from another measure different things, and the ranking reflects who scored rather than who fits.

Tie-breakers and who reviews the ranking

A tie-breaker is a criterion chosen in advance to order candidates with the same total score, for example recency of the core skill or proximity to the client's office. Writing it down before scoring stops the recruiter from settling ties by gut feeling.

The candidate ranking then needs a human review in two places: the top of the list, before anyone is contacted, and the profiles that failed a filter or scored lowest, to catch errors in the evidence. Profiles that score well but do not make the cut for this role are worth keeping in a talent pool for future roles rather than discarding.

Candidate scoring example: three profiles for one role

This candidate scoring example is invented to show the method. An agency recruiter is filling a senior data engineer role for a client in Barcelona. Must-haves: at least four years building data pipelines and working English. Weighted criteria: Spark or similar (40 %), cloud data stack (30 %), fintech experience (20 %), based in Barcelona or open to relocating (10 %). Scale 0–3.

Criterion Type Profile A Profile B Profile C
4+ years building data pipelines Filter Pass Pass Fail (2 years)
Working English Filter Pass Pass Pass
Spark or similar (40 %) Weight 3 2 —
Cloud data stack (30 %) Weight 2 3 —
Fintech experience (20 %) Weight 0 3 —
Barcelona or open to relocating (10 %) Weight 3 1 —
Weighted total (out of 3) Result 2.10 2.40 Out

Profile B ranks first even though Profile A has the strongest core skill, because the fintech criterion carries real weight for this client. Profile C is out on a filter, not on points, and the reviewer checks that the two years of pipeline work are not a profile that simply undersells itself.

How to rank candidates with AI in 2026

To rank candidates with AI, the recruiter writes the scoring criteria in plain language and the AI applies them to every profile in the list, returning the list ordered with the reasons behind each position. The AI does the reading at scale; the recruiter owns the criteria, checks a sample of scores and reviews every rejection before it becomes a decision.

Three rules keep AI ranking useful and defensible:

  • AI ranking uses criteria from the brief. The model should score skills, experience and location the recruiter defined, not infer fit from photos, names or other personal traits. More on the wider process in our guide to AI-powered recruitment selection.
  • AI ranking shows its reasons. A score without the evidence behind it cannot be reviewed, so the output should say which part of the profile met each criterion.
  • AI ranking does not reject on its own. Under Article 22(1) of the EU GDPR, a candidate has the right "not to be subject to a decision based solely on automated processing" with legal or similarly significant effects, so a person should sign off each rejection.

Under the EU AI Act, Regulation (EU) 2024/1689, AI systems used to analyse and filter job applications and to evaluate candidates are listed as high-risk in Annex III, point 4(a), and Regulation (EU) 2026/1744 set 2 December 2027 as the date their high-risk obligations apply; what that means in practice is covered in our guide to AI candidate screening and the EU AI Act.

How HeyTalent filters and ranks LinkedIn profiles

HeyTalent ranks LinkedIn profiles against the criteria the recruiter defines: AI variables written in plain language are applied to every profile in a search list, and the profiles that do not meet them are filtered out. In Agent Mode, the agent turns a brief into a search with advanced LinkedIn filters, reads every profile and orders them by the recruiter's requirements in a dynamic table.

Lists in sourcing are long enough to need this step: according to HeyTalent's data, the platform has analysed more than 100,000 unique LinkedIn profiles, and a typical search list gathers about 49 candidates. HeyTalent extracts those profiles in real time, so the experience being scored is the candidate's current one at the moment of the search.

The recruiter stays in charge of the result. HeyTalent's workflow ends with the person reviewing the list and the messages, adjusting what they want and only then launching the LinkedIn campaign. HeyTalent estimates that it has saved its users more than 1,000 hours of manual profile review.

HeyTalent ranks LinkedIn profiles, not CVs, and it does not score interviews: it does not extract CVs into candidate fields, and interview evaluation stays with the recruiter's own scorecard. HeyTalent does not make an AI ranking compliant by itself either: the EU AI Act duties of a deployer still fall on the agency or employer that uses the ranking.

Common candidate scoring mistakes

The most common candidate scoring mistakes are criteria that do not come from the role, years of experience used as a proxy for skill, and weights changed halfway through the search. Each one makes the ranking measure something other than fit for the job.

  • Criteria that do not come from the role. A criterion such as "worked at a well-known company" scores prestige, not ability, and it tends to repeat the profile of whoever filled the role last time.
  • Years of experience as a proxy for skill. Seven years in a role says how long someone stayed, not what they did; score the evidence of the skill instead.
  • Weights changed during the search. Adjusting weights after seeing a favourite profile turns the scorecard into a justification for a decision already made.
  • Scoring without anchors. Without written anchors, two recruiters scoring the same list produce two different rankings.
  • Unreviewed rejections. A filter applied to a badly written profile can drop a strong candidate; the review of the bottom of the list exists for that.

Several of these mistakes are forms of bias that structured scoring is meant to reduce, and our guide to unconscious bias in recruitment explains how they creep into screening decisions.

Join the new era of sourcing

Book a call today and start saving time.