Your Engagement Survey Is a Decent Turnover Predictor. It's Not Enough.
A synthetic-data walkthrough of building a turnover risk score — and why the metric most companies already have needs company, not replacement.
HR Analytics, Done Differently
Better decisions about people start with better questions.
Most HR analytics tells you what happened. I help you predict what will happen — and act on it with appropriate confidence.
Philosophy
The average employee doesn't exist.
Planning around point estimates — the average turnover rate, the typical engagement score — ignores the range of outcomes you actually face. I work with distributions, not just means. Uncertainty isn't a problem to eliminate; it's information to quantify.
01
Measure what matters, not what's easy
If it matters to your organization, it can be measured — perhaps not perfectly, but well enough to reduce uncertainty and improve decisions. That's the standard.
02
Prediction beats description
Dashboards tell you where you've been. Good models tell you where you're going. I build models designed to predict outcomes — turnover, performance, fit — not just report on them.
03
Calibrate your confidence
The goal isn't certainty — it's calibrated uncertainty. Knowing you're 70% confident in an outcome, and knowing what would change that, is more valuable than a false sense of precision.
Services
What I do
Workforce Prediction Models
Turnover risk, performance classification, promotion readiness — built and validated against your own data.
→Measurement Design
Define what you actually need to measure, build instruments that measure it, and validate that they do.
→Decision Analysis
Frame people decisions under uncertainty. Quantify what you know, identify what's worth learning, choose with clarity.
→Survey & Assessment
Rigorous psychometric design for employee surveys, assessments, and selection instruments.
→Insights
Recent thinking
Your "High Potential" Raters Probably Don't Agree With Each Other
A synthetic-data walkthrough of turning "high potential" into a measured construct — and what happened when two independent raters scored the same people.
Tenure and Performance Ratings Are Weak Predictors of Who Gets Promoted
A synthetic-data walkthrough of building a promotion-readiness model — and why the two inputs most companies rely on turned out to be the weakest ones in it.