Performance and Profit Prediction.
A job req value forecast model combined with a skills-based candidate matching algorithm that augmented recruiter intuition with data: optimizing effort and increasing margin.
Filling jobs on gut and post-it notes
Requisition triage ran on instinct and relationships. Recruiters worked the reqs that felt closest to the money and matched candidates from memory, spreadsheets, and post-it notes. The boolean search built into the ATS was high-friction and rarely used. Results were inconsistent: fill rates swung from 30% to 90% and averaged just over 50%, assignments ran 6 to 8 weeks, completion sat near 72%, and gross profit lagged.
The data to do better existed, but it was scattered across federated systems. New privacy rules such as GDPR and CCPA raised the stakes, recruiters were comfortable with their own methods, and talent-based prediction had never been attempted at this scale. Leadership was not sure it was even feasible.
Integrate, datafy, predict, embed
Integrate the whole picture
Client, market, and candidate data unified: firmographics and location intelligence, labor-market and talent-demand signals, and candidate experience, skills, and assignment history.
Datafy performance
Analyze 1M+ requisition, candidate, and assignment records to turn vague outcomes into measurable performance across 30+ criteria.
Predict what matters
Algorithms predict a requisition’s GP and a candidate’s performance, replacing intuition with an explainable, defensible score.
Embed where they work
Pilot an MVP in Tableau, then integrate predictions into the ATS for frictionless adoption, with a method to measure use and effectiveness.
From federated data to a scored queue
1Unify federated data
Client, market, and candidate data were unified from federated systems: firmographics and location intelligence, labor-market and talent-demand signals, and candidate experience, skills, and past assignments.
2Datafy performance
More than a million requisition, candidate, and assignment records were analyzed to turn vague outcomes into measurable, comparable performance signals.
3Build the 30+ criteria model
A data model of more than 30 criteria spanning job, client, candidate, and assignment dimensions, developed with regional data science and analytics teams.
4Predict GP and performance
Predictive algorithms estimate a requisition’s gross profit and a candidate’s likely performance, replacing intuition with an explainable score.
5Embed in the ATS
A Tableau MVP proved the concept, then predictions were integrated directly into the ATS for frictionless adoption, with a method to measure use and effectiveness.
6Govern for privacy
Designed around emerging privacy regulation and ethical-AI principles, replacing risky spreadsheet and post-it candidate lists with a governed, scalable system.
Which reqs to work, and who to place
A profit forecast was created for every open requisition and algorithmically tiered A, B, or C based on the quality and profitability of similar orders. Recruiters saw an improved recommendation and selection experience: try it for yourself.
Intuition, priced: $10.1M in gross profit
Within its first year at scale, algorithmically prioritized orders and candidate matching generated more than $10.1M in incremental gross profit, driven by an average four-point lift in fill rate, roughly one week longer assignment duration, and eight points higher assignment completion. Just as important, talent-based analytics earned credibility with the CEO, ELT, and Board, and recruiters adopted prediction because it lived inside the workflow they already used.
Whether you’re modernizing people analytics, deploying AI responsibly at scale, or sharpening product and market strategy, I’d be glad to talk.
The board, requisitions, candidates, and figures shown here are an illustrative reconstruction built from synthetic data. This work was completed while employed at a global workforce solutions firm; the production models and data remain the property of the company.