3800994 - How are jobs recommended by AI when uploading a CV / Resume via the Career site?

• A job is not being recommended to a Candidate despite some common skills?
• What guardrails are used by the system when recommending jobs?

SAP SuccessFactors Recruiting Marketing

1. Upload a candidate CV on the Career Site and relevant skills are extracted.
2. A job is posted and this has some common skills to the Candidate's extracted skill set.
3. Despite this, the job is not recommended 

https://community.sap.com/t5/human-capital-management-blog-posts-by-sap/how-does-the-career-site-job-recommendation-feature-work/ba-p/14127389

When a candidate uploads their CV, the system extracts skills, scores each job posting and returns recommendations in order of score. Two things affect the score:

• Primary skill matches are given a higher score than secondary skill matches
• Skills that appear commonly across many job postings (like "Communication" or "Teamwork") are given very little weight. Skills that are more specific and appear less frequently (like "Onboarding" or "Creative Problem Solving") are scored much higher

Before any job is shown to the candidate, two quality checks are applied:

• Check 1 – Score Drop-off Filter: The system compares all the job scores and if it finds a significant gap between the higher scoring jobs and the lower ones, everything below that gap is removed from the recommendations
• Check 2 – Minimum Skill % Filter:  Other remaining jobs must have at least 10% of its primary skills matched OR at least 25% of its secondary skills matched by the candidate's CV. If not, job is removed.

Both checks must pass for a job to be recommended.

A lower weight for skills that occur in many job postings can affect roles whose skill profile is dominated by broad or transversal capabilities, including some HR roles, internships, and junior positions. This is a consequence of the ranking model: common skills are less useful for distinguishing one job from another, whereas specific skills provide stronger differentiation. There is no documented SAP exception that disables the rarity weighting for HR, internship, or junior roles, and the feature does not assess seniority or proficiency. It extracts skills from the resume and matches them to the skills defined for the job. Therefore, a candidate can be substantively relevant to a role while still receiving a low recommendation score if the resume contains mainly common skills and does not contain enough of the job’s more specific skills. The practical guidance is to optimize the quality and classification of the requisition’s skill set, not to try to compensate by marking every skill as primary:

• Use Primary Skills for genuine must-have skills. Keep this set focused on the capabilities that are essential for performing the role. SAP defines primary skills as skills required for the job.
• Use Secondary Skills for complementary or nice-to-have capabilities. Common transversal skills can remain secondary when they support the role but are not decisive requirements. SAP defines secondary skills as complementary skills that enhance job performance but are not required.
• Include role-specific differentiators where they are genuinely relevant. For example, an HR role may have a small number of specific HR-process, HR-system, regulatory, or recruiting capabilities that distinguish it from other HR roles. These should be primary only when they are truly essential, and secondary when they are useful but not required. The examples here are illustrative; SAP does not prescribe a standard skill list for these role families.
• Make the required skills explicit in the job description and validate the extracted list. Recruiters can review AI-extracted skills, add or remove skills, and reclassify them as primary or secondary. This is the appropriate way to correct noise or ensure that important, role-specific skills are represented.
• Use realistic entry-level requirements for internships and junior roles. Since the matching logic is based on skill presence rather than proficiency or years of experience, the requisition should describe the skills expected at entry level and should not rely on seniority as an implicit matching signal.
• Avoid adding skills solely to improve ranking. Primary/secondary classification should reflect the actual job requirements. Adding too many broad or incidental skills may introduce noise and can make the resulting recommendation set less meaningful.

Best practice for these roles is a concise and accurate set of must-have primary skills, supported by secondary skills, with enough role-specific content to distinguish the requisition. 

• Additional details on skill extraction and Job Matching: How does the Career Site Job Recommendation feature work
• Using AI to extract skills: Extracting skills from job descriptions with AI

AI job recommendations, career site, not recommending relevant jobs, primary skills, secondary skills, score drop-off filter, minimum skill threshold, rarity weighting, skill matching, skills extraction, recruiting marketing, job ranking, recommendation guardrails , KBA , LOD-SF-RMK-INT , Int with RCM, Data Mapping, AwLi, Multilocation posting , How To