Candidate scoring in an ATS, explained

Candidate scoring is the ATS condensing what it knows about each candidate into comparable numbers: fit scores measure how well skills and experience match the role, flight-risk scores estimate the chance a candidate drops out, and readiness scores show how close someone is to making a move. Scores rank attention; humans still decide.

What is a fit score?

A fit score expresses how closely a candidate’s profile matches a specific role: required skills, experience depth, domain background, location and language constraints. The inputs come from structured data and from the parsed CV, which is why fit scoring is only as good as the CV parsing underneath it and the quality of the job requirements you wrote. A useful fit score explains itself: which requirements matched, which are missing, and which are uncertain. A bare “82” with no reasoning is a number you cannot defend to a hiring manager, a candidate, or a regulator.

Remember what fit measures: evidence of match on paper. It cannot see motivation, trajectory, or the skills someone never wrote down.

What does a flight-risk score mean?

Flight risk estimates the probability that a candidate slips away mid-process: goes quiet, accepts a counteroffer, or declines at the finish line. Typical signals include response latency, time stuck in a stage, the pace of your process compared with the market, and what the candidate has told you about competing conversations. The score is not a judgment of the person. It is an early-warning light for your process, and the correct response is action: tighten the timeline, communicate more, or revisit expectations before the resignation-letter moment arrives.

What is a readiness score?

Readiness asks a different question: how close is this person to actually moving? It matters most in talent pools and sourcing, where you choose whom to contact this week. Signals include recency of activity, openness cues, and career-stage patterns. Prioritizing high-readiness profiles makes outreach feel well-timed instead of random, and it protects the rest of the pool from being spammed into unsubscribing.

How are these scores calculated?

Every vendor does it differently, but the honest general description is: a blend of explicit rules and learned patterns applied to the structured data the system holds, parsed CVs, stage history, engagement events, and the role’s requirements. Two limits follow directly. First, scores inherit the flaws of their inputs: CVs exaggerate, parsers misread layouts, and stale data produces stale scores. Second, scores are estimates with error bars, even when the interface shows a confident integer. How the matching layer works in more depth is covered in how candidate matching works.

When you evaluate a vendor, ask two questions: which inputs feed each score, and can the interface show the reasoning behind an individual number. Vendors comfortable with both answers tend to have scoring you can defend later; vendors who answer with adjectives usually do not.

How should recruiters use candidate scores responsibly?

Scoring earns its place when it compresses reading time, and loses it the moment it starts making decisions nobody reviews. Six habits keep it on the right side of that line:

1. Scores order the queue; humans make the decisions. GDPR Article 22 restricts solely automated decisions with significant effects, and rejection is the obvious example.
2. Only act on scores you can explain. If the system cannot show its reasons, treat the number as a hint, not a fact.
3. Sample below the cutoff on a schedule. Ten minutes a week reading “low-scored” profiles catches systematic misses early.
4. Watch for proxy bias: employment gaps, school names, and postcodes can smuggle in patterns you never intended to reward.
5. Log overrides. If recruiters constantly override a score, the model is miscalibrated for your desk; that signal is valuable.
6. Say in your privacy notice that automated tools support evaluation, and be ready to explain an outcome to a candidate who asks.

What scoring looks like in Recruitifly

Recruitifly scores every pipeline candidate on fit, flight risk, and readiness, and puts the numbers on the board where prioritization actually happens, next to the stage, not buried in a report. Fly, the built-in assistant, can rank a shortlist, explain why a candidate scored the way they did, and propose next steps, with every action behind a propose-then-confirm gate so a human approves anything that changes the record. You can see how that fits the wider workflow on the Fly page.

Recruitifly is in private beta. If you want a pipeline that tells you where to look first without deciding for you, talk to us and join the beta.