Success Metrics for LinkedIn's Profile-to-Job Match Percentage Feature

LinkedIn has launched a feature for job seekers called the "Match Calculator." When a member opens a job posting, the feature shows a percentage estimating how well their profile fits that job's description. You are the product manager responsible for it. How would you measure whether the feature is a success? Cover what the feature is meant to achieve, the primary metric you would use to judge it, the supporting and guardrail metrics around that metric, and how you would show that any movement was caused by the feature. ```hint Start from the decision it changes The percentage is only valuable if it changes what a job seeker does next on that posting. Name that decision, and what a better version of it means for the seeker and for the recruiter reading the application, before you pick any metric. ``` ### Clarifying Questions - Is the match percentage shown to every job seeker, or only to a subset such as paying subscribers or a test group? - Do recruiters see the same score on incoming applications, or is it visible only to the applicant? - Does the feature explain the score, for example by listing the requirements the profile is missing, or is it a bare number? - Has it launched to everyone already, or are we deciding whether to expand a limited rollout? ### What a Strong Answer Covers - A goal that serves both sides of LinkedIn's jobs marketplace: applicants choosing where to apply and recruiters reviewing who applies - One primary metric tied to better-fitting applications and their outcomes, rather than to views of or taps on the score - A funnel from seeing the score to applying, plus downstream recruiter signals such as replies, shortlisting and interview requests - Guardrails for the specific risks of a fit score: fewer applications per job, qualified people discouraged by a low number, profile keyword stuffing, and uneven accuracy across groups of members - An experiment design that can attribute impact, and how competition between applicants for the same jobs complicates it ### Follow-up Questions - In the experiment, applications per job seeker fall while recruiter reply rates rise. Would you call the feature a success, and what would you check before deciding? - How would you tell whether the percentage is accurate, separately from whether people use it? - How would you detect that members with low scores stopped applying to jobs they would have been hired for?