Resume scoring rubric — how the 100-point score works

Resume AI scores every resume on a 7-rule rubric out of 100 points. Same number every time you look at the same resume — deterministic, no AI grading the score itself. The full breakdown is visible to Pro+ / Lifetime; Free + Hour-Pack tiers see only the top-line score.

The 7 rules

Rule	Points	What it measures
Contact Info	10	Full name, email, phone, location, LinkedIn. All 5 fields present = 10. Missing fields cost ~2 points each.
Experience	25	Number of roles + dates + employer names + actual bullet content (not "responsible for…" template leftovers). Largest weight — this is what interviewers actually scan.
Education	10	Degree(s), institution, graduation year. Bonus for relevant coursework if listed.
Skills	10	A distinct Skills section with concrete tools / languages / frameworks. Lists of buzzwords without specifics score lower.
Content Quality	20	Bullet structure (action verbs, quantified results, specifics). Generic bullets ("worked on team," "responsible for X") cost points; specific bullets ("shipped Y feature to 12M users, cut latency 30%") gain them.
Page Fullness	15	One full page = good. Half a page = under-claiming. Two+ pages = signal of poor editing. Sweet spot is 0.95-1.05 pages.
Originality	10	Penalizes detection of the default-template strings (e.g., "Alex Johnson," "Acme Technologies," "responsible for software development"). If you didn't edit out the dummy placeholders, this drops fast.
Total	100

Why rules and not AI

Three reasons the score uses deterministic rules instead of LLM-graded subjective scoring:

1. Instant. Each score recalculation is sub-50 ms (vs ~2 seconds for an LLM round-trip).
2. Deterministic. The same resume always produces the same number. You can A/B test "what if I rewrite this bullet" and trust the delta.
3. Predictable. AI scoring drifts — same resume scores 73 today, 78 tomorrow because the model was tuned slightly differently. Rules don't drift.

Full design memo: engineering/resume_score_rubric.md (internal).

What a "good" score is

• 70-79 — solid resume, hits most rules. Most people start here.
• 80-89 — strong resume, clearly edited, quantified. Above-average.
• 90+ — exceptional. Rare on first upload; usually requires deliberate iteration with AI rewrite.
• <70 — something's missing: probably Contact Info gaps, dummy text not edited out, or Content Quality bullets that need quantification.

How to improve your score

Each rule has a hint pointing at the specific issue. Click the rule on /resumes → Score Panel → see the hint text. Common fixes:

• Contact Info -8 → "Phone or LinkedIn missing" — add the field
• Content Quality -12 → "Most bullets lack quantified results" — rewrite bullets with numbers (use AI Rewrite for help)
• Originality -10 → "Template defaults detected: 'Alex Johnson', 'Acme Technologies'" — replace dummy text
• Page Fullness -7 → "Resume is 0.6 pages" — add a Projects section or expand bullets

The score is one signal, not the signal

A 95 resume can still fail interviews. A 75 resume can land FAANG. The rubric grades the resume artifact; it doesn't grade you. Use the score to find quick wins (template leftovers, missing contact info) — not as a hiring oracle.

Score visibility per plan

Plan	What you see
Free	No score visible — locked entirely
Hour Pack	Top-line score only ("78/100")
Pro / Pro+ / Lifetime	Full per-rule breakdown + hints

Where it lives in code

apps/api/domains/scoring/resume.py — the 7 rules, byte-identical port from the legacy vanilla-HTML resume builder. Frontend Score Panel: apps/web/app/resumes/_components/score-panel.tsx.

Still stuck?

Help & Feedback → Contact us with category "How do I…?" if your score doesn't match what you expected, OR if a rule's hint is unclear.

Video walkthrough

Coming soon.