Every rule behind the number
This is the whole scoring model, not a summary of it: 17 rules across three layers, each with a fixed weight and a threshold that decides pass from fail. The same file and the same posting always produce the same score.
Layers add up
Formatting 30 + keyword coverage 40 + content quality 30 = 100.
No posting, no keyword layer
Without a posting, the keyword layer is skipped and the remaining 60 points are rescaled to 100.
Whole points only
Each rule is rounded to a whole point, and a layer is the sum of its rules, so the ledger always adds up.
Verdicts have a floor
75 and above is “strong”, 58–74 is “okay”, below 58 is “weak”. A strong total drops to okay if the name or email cannot be read, or if a table or multi-column layout fails.
Formatting
Nine checks that decide whether a parser can read the file at all. A layout failure costs points even when the writing is good.
Contact block
A name, an email address, and a phone number that survive parsing. Icons, text boxes, and graphics do not count — the values have to exist as text in the document body.
Two points per field that reads. All three present passes; two is a warning; a missing name or email also caps the whole score below “strong”.
Tables
Any table in the document, including a table used only for layout.
Zero if a table is found. A failing table also prevents a “strong” verdict, even when the total reaches 75.
Why tables break parsing →Columns and text boxes
More than one column of text, including side-by-side blocks and text boxes.
Zero unless the flow is single-column. A failing column check also prevents a “strong” verdict, because reading order is not guaranteed.
Standard section headings
Experience, Education, and Skills under their conventional headings.
Four points split evenly across the three headings. All three is a pass, one missing is a warning, two or more missing fails.
Length
Page count and word count after parsing.
Full points for two pages or fewer; zero past page two.
Images and graphics
Photos, icons, logo marks, skill bars, and any other image object.
One of two points when any image is detected. Images carry no text, so a fact that lives only in a graphic is lost.
Non-standard fonts
Typefaces outside the standard set — Calibri, Arial, Georgia, Times, Verdana, and similar. Internal font ids are ignored.
One of two points when an unrecognized typeface appears; decorative faces often fail to map characters correctly.
Date format consistency
Whether every role uses one date style.
Full points when one style is used, or when no dates were found; zero when styles are mixed. “Jan 2021 – Mar 2023” counts as one style.
Keyword coverage
Three checks against one specific posting. The layer runs only when at least 40 characters of a posting are supplied; without one it is skipped and the other 60 points are rescaled to 100.
Core skill terms
Named skills from the posting that also appear in the resume. Up to 22 terms are tracked, and common synonyms count as a match.
Worth 24 when the posting also yields recurring phrases, 34 when it does not. 90% present passes; 60% is a warning.
How keyword coverage is counted →Posting-specific phrasing
Two- and three-word phrases that recur at least twice in the posting, up to eight of them. This is the posting's own wording, not a generic synonym.
Worth 10 alongside dictionary skills, 34 when the posting yields no dictionary terms. The same 90% / 60% thresholds apply.
Title alignment
Whether the posting's title is reflected in the resume: at least 70% of the title's meaningful words appear in the text.
Full 6 points when aligned; 3 points and a warning when not.
Content quality
Five checks on the bullets a screener actually reads. Experience, projects, and volunteer sections are scored first; if those are empty, every bullet is used. A file with no bullets scores zero for this layer.
Bullets without a measurable result
Bullets with no number: no volume, percentage, time saved, budget owned, or error rate.
Up to 20% of bullets may lack a number for a pass; up to 50% is a warning. Credit falls in proportion to the share without one.
Weak openers
Bullets that open with “responsible for”, “helped with”, “worked on”, and similar duty phrases. Common Chinese duty openers such as “负责” and “协助” are treated the same way.
A pass at 10% or less of bullets, a warning at 40% or less, partial credit in proportion.
How to rewrite a duty bullet →Bullets over two lines
Bullets longer than two rendered lines.
A pass at 10% or less, a warning at 35% or less, partial credit in proportion.
Repeated opening verbs
One verb opening three or more bullets inside the same role.
No repetition passes, one repeated verb warns, two or more fail. Credit falls at twice the rate of the share of repeated verbs.
Duties restated instead of outcomes
Bullets that describe a duty with no outcome, including any weak opener that also carries no number.
A pass at 10% or less, a warning at 40% or less, partial credit in proportion.
It does not reproduce any specific vendor's parser, and it does not predict whether you get an interview. It measures the mechanical failures that are cheap to fix before a human reads the file — nothing else.
The free score shows every failing rule with its weight and the exact fix. No account, no card.