Below, I provide a brief overview of the project's background and results. If you're interested in understanding directly how I arrived at this solution, please skip to My Role & Process section.
A crucial aspect of our job-seeking platform's growth is the users' propensity to return. However, we've observed a limited number of returning users. This trend results in higher costs to continuously acquire new users, with few of them revisiting our platform. Upon investigation and feedback collection, it appears that a primary concern is the lack of job recommendations that align with their profiles.
Based on our findings, we've developed the following hypotheses:
Hypothesis 1: Over-reliance on specific keyword matching.
The previous matching system primarily depended on keyword similarities between job titles and a talent's prior experiences. This might have caused a rigid and reduced number of recommendations.
Hypothesis 2: Ignoring users' experience levels in recommendations.
While the system did match the number of years of experience between users and job role requirements, it failed to differentiate and match specific roles. For instance, pairing 'manager' with 'staff' or 'staff' with 'senior lead' were both considered matches.
Hypothesis 3: Salary range not factored into recommendations.
There were instances where users provided their expected salary, but the algorithm didn't factor this into its recommendations.
Upon forming our hypotheses, we initiated the ideation process:
match?matching criteria.Improved job matching aligned with users' profiles resulted in a 53% surge in job applications. This refinement also assisted recruiters in identifying more fitting candidates for their job listings.
My Profile
Jobs Curated for Me