What an ATS match score actually measures, and how to read yours

An applicant tracking system (ATS) is the software a company uses to collect applications. Most of them can rank or filter candidates against the posting, and recruiters often work from that ranking before a person reads anything. An “ATS match score” is an estimate of where you would land in that first screen.

No tool outside the company can see its real ranking. What a score can do is apply the same kind of rules a keyword screen uses, consistently, so you know before applying whether the obvious gaps are real or just missing words. This post explains what the Lazee score counts, so the number means something when you read it.

What goes into the score

The score is 0 to 100 with a breakdown by section, and the breakdown is the useful part:

• Skills. Which of the posting’s skills you list, split into matched, related and missing. Required skills count more than nice-to-have ones.
• Years of experience. The posting’s stated years against the span of your work history.
• Role fit. Whether your titles belong to the same role family and seniority as the posting. Full stack is close to backend and far from sales.
• Education. A degree requirement, when the posting has one.
• Location, visa and work setup. Only when the posting states them, and only against the job preferences you filled in.

The rules that change the number

A naive keyword count rewards stuffing. These rules are there so the score reflects how a posting is actually written:

• Sections matter. Skills and years mentioned in “About us” or “Benefits” do not count. Only the requirements and responsibilities do.
• Sentences matter. “Good to have: GraphQL” makes GraphQL a soft requirement. “No PHP experience needed” counts for nothing.
• A skill in the job title counts double. A “React Developer” posting is mostly about React.
• Alternatives are one requirement. “AWS, GCP or Azure” is satisfied by any one of them.
• Implied skills count. PostgreSQL implies SQL, Next.js implies React.
• Related skills earn half credit. If the posting wants Vue and you list React, or wants GCP and you list AWS, that is a related match, not a miss.
• Unknown tools still count. A product the engine has never heard of is treated as a skill when the posting lists it after “experience with”.
• Seniority implies years. A “Senior” or “Staff” title with no stated years assumes the usual number, at half weight, and the breakdown says so.
• The company’s own name is never a skill.

Hard limits cap the score

Some gaps are not about skills. If the job must be done from the US and your preferences say you need sponsorship and are not there, or the posting is on-site in a city you have not listed, the score shows a red warning and is capped, however good the skills match is. That is deliberate: a recruiter would stop at the same line. A “Make this score more accurate” box lists the preferences you left blank, because a blank is treated as unknown, not as a clash.

Where to run it

The free score runs offline inside the extension against your cached profile, so nothing is sent anywhere. It needs a minimally complete profile: at least three skills and one work experience or project. You can run it from:

• the ATS Match tab in the Lazee modal on any application form;
• the extension popup, which scores the job open in the current tab;
• the ATS match button the extension adds to listings on LinkedIn, Wellfound, Work at a Startup, Superteam, Google Careers and Recruitee career sites;
• the ATS match button on every listing of the job board.

The AI rescore

Pro users (or free users spending one Pro Pass) can ask for an AI second opinion on top of the offline score: a recruiter-style read with strengths, gaps, concrete profile fixes and a list of skills to add if you actually have them. It costs 6 credits on Lazee AI and 0 with your own key. Only fit-relevant profile data goes to the model (skills, titles, dates, descriptions, projects, degrees, languages, preferences), never contact details.

How to raise a low score honestly

1. List the skills you already use. Most low scores come from a thin skills list, not a thin background. If you work with Docker every week and it is not on your profile, the screen cannot know.
2. Fill in job preferences. Work authorization, locations and languages turn unknowns into matches, and stop the warning on jobs you are eligible for.
3. Use the matching resume variant. The score reads your profile, but the recruiter reads the PDF; keep them telling the same story.
4. Do not add what you do not have. A padded profile passes the screen and fails the interview, which costs more time than the application saved.