A low matching score in CV-to-JD matching but a high score in JD-to-CV matching (or vice versa) is a known behavior in systems like RChilli's Search & Match API, and it often results from directional scoring logic. Here's a detailed breakdown to help you understand why it happens and what to check:
Understanding the Direction of Matching
| Matching Direction | Query Source | What Is Compared | Focus |
|---|---|---|---|
| JD-to-CV Matching | Job Description (JD) | Compares required job criteria (skills, experience, domain, etc.) | How well resumes meet the job requirements |
| CV-to-JD Matching | Candidate Resume (CV) | Matches the candidate’s profile to all indexed job descriptions | How well jobs align with the candidate’s profile |
Why scores may differ:
These two directions interpret relevance differently based on missing data, weightage, and query structure.
Why the Scores Can Differ
| Reason | Explanation |
|---|---|
| Scoring Weights Are Directional | RChilli's default scoring logic gives different emphasis depending on the query origin. JD-to-CV may weigh mandatory skills, education, or location more heavily, while CV-to-JD may weigh experience, past roles, or skill coverage more. |
| Query Analyzer Behavior | When a JD is the query, the system uses keyword tagging (e.g., must-have vs. nice-to-have). When a CV is the query, it may interpret fewer intent-based signals—especially if the resume lacks structure or clarity. |
| Resume Lacks Keyword Density or Alignment | The candidate's resume might be missing synonyms or taxonomy terms that are used in the JD. JD-to-CV match still works well due to taxonomy expansion, but CV-to-JD lacks precision. |
| Field Availability & Enrichment | JDs are often short and focused (specific roles, domains, skills). CVs may be broad and generic, with less structured or missing fields (e.g., education not detailed, skills scattered). This causes misalignment during score computation when CV is the query. |
How to Improve CV-to-JD Scores
| Action | Recommendation |
|---|---|
| Use ParseAndIndex with Taxonomy | Ensure resumes are enriched with RChilli’s taxonomy and aliases. |
| Enable API Settings | Add dynamic settings like reqskillsdrill, reqtaxonomy, degreeorder for more structured output. |
| Ensure Field Completeness | Check that resumes have full details: Skills, Experience, Education, Certifications. |
| Review Weight Configuration | Adjust scoring weightage (e.g., more weight to skill match in CV-initiated queries). |
Example Scenario
| Case | JD (Job Description) | Resume (CV) | JD→CV Score | CV→JD Score |
|---|---|---|---|---|
| Job Title | Looking for "Full Stack Developer, React + Node" | Resume missing exact phrase "Full Stack Developer" | High (due to alias match like "JavaScript Developer") | Low (resume lacked role clarity or skill hierarchy) |
Diagnostic Tip
Use the Search & Match Query Analyzer to compare:
-
What keywords are extracted from the JD
-
What fields are populated in the CV
-
Which fields contribute to the score
For more detailed information, visit RChilli Search and Match API Overview.
Need Help?
Contact support@rchilli.com for personalized scoring configuration.
Comments
0 comments
Please sign in to leave a comment.