Methodology
Transparent documentation of how ApplyIntelligent measures results, calculates scores, and reports user outcomes. We believe in complete transparency so you can judge our claims and understand our limitations.
Why this page exists
This page explains how ApplyIntelligent measures results (scores, improvements, and user-reported outcomes). It's here so you can judge our claims, replicate our process, and understand the limits of our data.
What we measure
1) ATS Score (RQS: Resume Quality Score)
A composite score estimating how well a resume will parse and match a job description.
- • Structure & parsing readiness
- • Content relevance (keyword alignment)
- • Clarity & readability
- • Formatting hygiene
2) Keyword Gap
Top skills/terms a job description emphasizes that the resume under-represents or omits.
3) Bullet Strength
Change in quality of resume bullets (specificity, measurable impact, action verbs).
4) Application Velocity
Self-reported minutes saved per tailored application.
Definitions & formulas
ATS Score Calculation (0–100)
Keyword Gap
TF–IDF + embedding similarity of JD vs. resume, mapped to a skills ontology, surfacing Top N missing skills.
Bullet Strength (0–5)
Human-rater rubric (blind): Specificity (0–2), Impact (0–2), Clarity (0–1). Plus acceptance rate tracking.
Data sources
Product Telemetry
Opt-in anonymized metrics (scores, rewrites, exports). No content storage without consent.
User Surveys
Opt-in surveys for interview outcomes, time saved, and satisfaction ratings.
Public Job Descriptions
Used to refine skills ontology and benchmark tests across industries.
Privacy: We aggregate analytics and remove identifiers. You can delete your data anytime from account settings.
Measurement protocols
ATS Score Calibration
- • Weekly A/A checks to verify scoring stability
- • Benchmark set of synthetic + real resumes/JDs across roles
- • Manual spot checks where parsers disagree
Pre–Post Design
When users tailor a resume for a specific JD, we track: ΔATS, ΔBullets, Time to Tailor (median minutes)
A/B Experiments
New features tested against control (random assignment). Report differences as median deltas with 95% CIs.
What we will—and will not—claim
✅ We WILL report:
- • Median improvements with confidence intervals
- • Clear labeling of self-reported metrics
- • Cohort sizes and methodology details
- • Example: "median +14 ATS points (95% CI: +12 to +16) for PM cohort, n=412"
❌ We will NOT claim:
- • Causality for interviews or job offers
- • Vanity numbers without clear methodology
- • Guaranteed outcomes or success rates
- • Results from cohorts with n < 100
Limitations & bias
Known Limitations
- • Self-reporting bias: Users may over/underestimate interviews and time saved
- • Market conditions: Seasonality, layoffs, and hiring booms affect outcomes
- • Role diversity: Improvements differ significantly by discipline and seniority
- • Sample limitations: We don't publish cohorts with n < 100
- • Parsing vs. hiring: Great ATS score helps visibility, doesn't guarantee offers
How you can verify our claims
Export Demo Report
Get ATS score, keyword gaps, and rewritten bullets to verify consistency
Replication Kit
Use sample JD/resume pairs with expected scores to check our methods
Track Your Results
Monitor your own 10-application baseline vs. tailored results
Versioning & change log
- v1.0 (January 2025): Initial public methodology, added bootstrapped confidence intervals, set n≥100 threshold for reporting
- Future updates will be logged here with specific metric or model changes
Contact & questions
Questions about our methodology?
Reach out for clarifications, replication kit access, or methodology discussions.