How Shortlist scores a job

Short answerShortlist lists a posting's requirements, checks each against your profile and gives it a value (met 1, inferred 0.6, partly met between 0.2 and 0.6, always below a requirement whose tools are all listed, unknown 0.3, no evidence 0). The score is the weighted average, with required items counting three times a nice-to-have, then capped if a required or core item has no evidence. Named tools and years of experience are checked by code, not the AI.

A match score is only useful if you can see why it is what it is. This is the complete method Shortlist uses, including the parts that are not done by AI. The score measures how well a posting's requirements are evidenced by your profile. It is not a chance of being hired.

The profile is the only source

Every score reads your profile in Shortlist (Me → Profile), not your resume file. The profile is the single list of facts about you that scores, resumes and form answers all use. If your resume names skills the profile does not, Shortlist tells you and offers to bring them in, so the two do not drift apart. A note you add to a specific job ("I have used Spring Boot") counts as profile for that job.

Step 1: list the requirements

The posting is read once and turned into 4 to 24 requirements. Each one is marked:

required, preferred or nice to have, from the posting's own wording;
as a skill, experience, credential, eligibility or other requirement;
core if the job is about it (at most four per posting).

Duties ("you will design services…") are dropped; only what the candidate must bring is scored. This step uses AI, and its result is shared by everyone who scores the same posting, so the same job is always measured against the same list.

Step 2: check each requirement

The AI reads your profile against each requirement and quotes the line it relied on. Then fixed rules in code check the parts that should not be left to judgement, and the AI cannot overrule them:

Years of experience and location are calculated from your profile's dates, city and relocation choices. Only a relocation preference written as free text is read by the AI, in one narrow question.
Named tools are checked one by one. A requirement like "Spring Boot, Spring Security, OAuth2/SSO" is three things, and each is looked up in your profile under its known spellings (OAuth2 is OAuth 2.0; SSO is single sign-on). For each, Shortlist records where it was found ("Experience: your employer", "Projects", "Skills") and the line.
"/" and "or" mean a choice. "Oracle/SQL Server/MySQL" needs any one. A comma list needs all of them, unless it ends in "or".
Related is not the same. Spring Boot is not Spring Security, PostgreSQL is not Oracle, Docker is not Kubernetes. Related experience can make a requirement inferred, never met.
Plain-language parts that name no tool ("scheduled jobs at scale") are judged by the AI from your experience.

Step 3: give each requirement a value

Status	Meaning	Value
Met	Shown in your profile: in work, a project, or listed with use	1
Inferred	Related experience, or listed without depth shown	0.6
Partly met	Some of its named tools or parts shown	0.2 + 0.4 × the share shown
Unknown	Not stated either way	0.3
No evidence	Nothing in your profile shows it	0

A partly met requirement always counts for less than one whose named tools are all in your profile: three of four tools shown is worth 0.5, all four listed is worth 0.6.

"No evidence" means your profile does not show it, not that you lack it. Add it to your profile if it is true and the score updates.

Step 4: the score

Each requirement is weighted: required 3, preferred 2, nice to have 1. The score is the weighted average of the values, out of 100. Then two caps apply:

A required item with no evidence keeps the score below Strong (at most 74).
A core item with no evidence, or a dealbreaker such as a required clearance you do not hold, keeps it below Potential (at most 49).

Bands: Strong 75 and above, Potential 50 to 74, Weak below 50.

A worked example

A backend posting with five requirements, scored against a profile that shows Java, Spring Boot, OAuth2 and SQL:

Requirement	Level	Status	Weight × value
3+ years of Java backend (core)	required	met	3 × 1 = 3
Spring Boot, Spring Security, OAuth2/SSO	required	partly met: 2 of 3 shown	3 × 0.47 = 1.4
SQL/PL-SQL	required	partly met: 1 of 2 shown	3 × 0.4 = 1.2
Kubernetes	preferred	no evidence	2 × 0 = 0
Prompt engineering for code generation	nice to have	inferred	1 × 0.6 = 0.6

Total 6.2 out of a possible 12, so the score is 52, Potential. No required item is wholly missing, so no cap applies. The requirement list shows "Partly: Spring Security not shown" and "Partly: PL/SQL not shown", which tells you exactly what to add to your profile, if it is true, or to ask about in an interview.

Had the profile shown no Java at all, that core requirement would have no evidence and the score would be capped at 49, Weak, however well the rest matched.

Same inputs, same score

Results are stored per posting, profile and note. Scoring the same job again with the same profile returns the same result; changing the profile or the note re-scores it. The AI step can vary by a few points between fresh runs, which is why a result is kept rather than recalculated each time you open the job.

FAQ

Is the score my chance of getting the job?

No. It shows how well the posting's requirements are evidenced by your profile. Use it to decide where to spend your time, then read the strong matches yourself.

Why does a skill I have show as "no evidence"?

Because your profile does not mention it, or mentions a related tool instead. Add it to your profile, ideally in the experience or project where you used it, and score the job again.

Why is a listed skill only "inferred"?

A skill that appears only in a skills list, without any work or project that shows it, counts as inferred (0.6). Mention it where you used it and it counts as met.

Does the AI decide the score?

No. The AI lists the requirements and judges plain-language experience. Named tools, years, location, the values, weights and caps are all fixed rules in code.

Can I check my resume the same way?

The free ATS checker reads a resume the way a tracking system does and scores it on a posting's keywords, without an account.