A fit score is a claim about your graph, so you should be able to check how it was reached. This is the whole method.
A posting is split into separate requirements, kept close to its own wording, and each one is typed as one of three things.
A years figure is recorded only when the posting states a number. Five or more years is five. Extensive experience is nothing, because it is not a number.
Every requirement is put against your graph: positions, accomplishments, projects, skills, publications, certifications, education. Each comes back one of three ways, with the reasoning written out and the single strongest piece of evidence named.
metSomething specific in your graph supports it. On the run shown on the home page, “developing immunoassays and protein detection methods” came back met.
partialThe nearest evidence is adjacent rather than direct: the same method in a different setting, the same tool without the stated depth, or the right work without the years asked for. On that same run, “PhD plus 5+ years of post-doctoral industry experience” came back partial.
gapYour graph is silent on it, or contradicts it. A gap is written down rather than smoothed over, and nothing generated afterwards claims it.
One rule cuts across all three. Where a requirement names an environment, a scale or an accountability, that is part of the requirement, and evidence from an easier one is partial at best. Regulated work is not met by an unaudited analogy. Running a service on call is not met by having built one. Owning a budget is not met by contributing to one. A license a role requires is met by holding it and by nothing else. The rule reads the same in every field, because a standard that is strict in one trade and loose in another flatters some people and punishes others.
The model judges the evidence. The arithmetic is done in code, which is why the same judgements always produce the same number.
Each judgement carries an evidence strength from 0 to 1. Each requirement carries a weight from its type: a must-have counts 3, a responsibility 1.5, a nice-to-have 1. The score is the weighted average of those strengths, expressed as a percentage and rounded.
Nothing else enters it. There is no adjustment for how the posting is written, how many requirements it happens to list, or how your CV was formatted.
Then two caps, because an average can hide the one thing that decides an application. A single missing must-have takes a strong down to moderate. Three missing must-haves, or gaps in more than a third of them, take a strong or a moderate down to stretch.
Each verdict carries a recommendation. Strong is apply. Moderate is apply when no must-have is missing, and reframe when one is. Stretch is reframe when there are at least as many partials as gaps, and develop when gaps outnumber partials. Poor is skip.
It can still be wrong. It reads what is in your graph, so a thin graph, or a CV read badly, produces a thin judgement. Every record is editable, and correcting one changes the next run.
A score is a reading of your graph against one posting. It is not a prediction about what a hiring manager will do.