Let’s Fix How We Hire Hourly Workers

High turnover among hourly workers gets treated as an unavoidable cost of doing business. It isn't. The real breakdown happens at the decision-making stage: hiring managers get plenty of applicants but too little time to evaluate them properly, and end up making rushed, low-confidence choices that lead to quick turnover and a repeat of the same hiring cycle within months.

Recruitment automation is often pitched as the fix, but automating a broken process just produces bad hires faster. The better approach is building systems around what actually predicts quality hires at a given company, rather than assuming automation alone solves anything.

Key points covered:

• Protected characteristics like age, gender, ethnicity, and even names should be excluded from any hiring algorithm, both for legal reasons and because they carry no predictive value.
• Machine learning models should be trained on a company's own applicant and post-hire data rather than a generic off-the-shelf model, since what predicts success varies by employer and by role.
• Some intuitive data points, such as whether a candidate is currently employed, often turn out to matter far less than expected, while other factors, like commute time relative to wage, can strongly predict early turnover.
• Tenure is a reasonable starting proxy for quality of hire, though not a perfect one, and should be paired with other post-hire performance data.
• Removing demographic data from the algorithm reduces bias compared with purely human decision-making, and diversity outcomes can be monitored and adjusted within the system rather than left to chance.
• The goal is matching candidates to the traits shared by a company's best current employees, improving the odds of a good hire rather than guaranteeing one.

The underlying argument is that treating high turnover as inevitable wastes money and burns out hiring managers who are already stretched thin. Confidence in hiring decisions can be built with the right data, not just gut instinct under time pressure.