Matching smarter
A little over a decade ago, I participated in a project for a client, a large recruitment agency, aiming to enhance the matching feature in their ATS to boost their recruiters' efficiency. The goal was to reduce the time taken to fulfill assignment requests from clients to the shortest possible ‘shelf time.’ Analysis revealed that the shorter the turnaround time, the more requests a recruiter could handle each day, either closing deals successfully or dismissing them when unsuitable.
Our team prioritized improving positive outcomes. To identify how to optimize deal closures, we examined historical assignment data from both the back-office system and the ATS. While data analysis isn't my forte, reviewing the findings was enlightening for me and other business-oriented team members. We discovered patterns that seemed intuitive in hindsight. For instance, the recruiter's prior familiarity with the candidate was the primary factor contributing to successful and swift deal closures. Knowing a candidate from previous placements significantly expedited the matchmaking process. Despite working for a job-matching software company, this insight wasn't commonly considered in matching algorithms then, and it's still not frequently utilized today.
Another crucial point was the amount of time a candidate spent on customer assignments, particularly the duration per assignment. Candidates who worked longer hours per assignment were more appealing, as these assignments directly correlated with more closed deals per recruiter over a longer period. Surveys conducted among current and former candidates revealed that many temporary workers transitioned to permanent roles after a certain period or number of assignments. This led us to establish the average number of assignments per job type/profession, considering data from external sources and the client's ATS. This information influenced matching decisions, prioritizing candidates with a history closer to the average for their job type.
Additionally, we identified other important data points to refine our algorithms, all unrelated to job requirements. This approach made sense since successfully placed candidates typically matched job requirements in terms of skills, location, experience, and certifications. Consequently, we opted not to change the algorithm for job requirements.
When we tested the new matching tool with a group of recruiters, we were pleased to find that the results were well-received, requiring minimal feedback. Recruiters found the results more familiar and recognizable, with known candidates appearing higher in the list. Upon rollout, the results exceeded our client's expectations, doubling efficiency compared to the financial targets set for the project. Recruiters appreciated the tool as it facilitated more deal closures and improved relationships with candidates and clients.
Today, recruiters are content when they find suitable candidates and actively pursue job opportunities. However, the insights gained from our project could prove valuable in a changing market with fewer candidates.
Feel free to share any unexpected data points or criteria that have helped you close deals, unless these are trade secrets, of course ;-)