People analytics & psychometrics

Where organizations get this wrong

• An engagement survey that has run for years without anyone checking whether its scales are reliable or its items mean the same thing to different groups.
• A hiring assessment chosen on a vendor's word, with no local validity evidence and no adverse impact analysis on file.
• Dashboards full of percentages that nobody can connect to a decision, because the underlying measure was never designed to support one.
• A "driver analysis" that is a correlation table with an arrow drawn on it.

Each of these is fixable, and each is cheaper to fix before the data are collected than after.

What I do

Survey science and employee listening

Designing instruments that measure the construct you actually care about: item writing, response scales, pilot testing, factor structure, reliability, and measurement invariance across groups so comparisons are legitimate. Weighting and sampling where the population is uneven. Driver analysis that models the outcome rather than describing it, and reporting that gets used.

Assessment development and validation

Building selection and development instruments end to end: job analysis and competency definition, item development, item response theory and classical item analysis, convergent and construct validity, criterion-related validity where criteria exist, and the technical documentation that makes the whole thing defensible.

Fairness, adverse impact, and defensibility

Differential item functioning, measurement invariance, and adverse impact analysis against EEOC and UGESP expectations. This is increasingly what buyers of assessment need most, and increasingly what regulators ask about first.

Translation

The part that decides whether any of it matters. I explain what a result does and does not support, in plain language, to people who did not train in measurement and should not have to.

Areas I have measured

• Emotional intelligence. Perceiving, using, understanding, and regulating emotion, and the harder problem of measuring it without relying on self-flattery.
• Cognitive abilities. Reasoning, problem solving, and learning ability, where item difficulty and fairness analysis carry most of the weight.
• Job skills. Task-specific and technical capability, anchored to job analysis so the test measures the job rather than a proxy for it.
• Personality and work styles. Trait structure, response distortion, and the difference between what predicts performance and what merely describes someone.
• Individual strengths. Development-oriented measurement, where the goal is useful self-knowledge rather than a pass or fail decision.

How this shows up in practice

• Eight assessments in seven months. At SciRec, from underlying theory through item development to full psychometric validation, including reliability, factor structure, validity evidence, and bias testing.
• An organizational culture measure. The Ethical Culture Scale, a 20-item, 5-factor instrument developed and validated at Ethical Systems, released free for academic use.
• Speak-up behavior at a large corporation. A study designed to identify what drives employees to report unethical behavior or stay silent, run inside a live organization.
• Pre-employment adaptability assessment. Led validation for a three-person build: reliability, factor structure, criterion-related validity.

More selected work and publications.

Methods and tools

Item response theory, exploratory and confirmatory factor analysis, structural equation modeling, reliability and generalizability, DIF and measurement invariance, mixed-effects and regression models, segmentation and cluster analysis, conjoint and MaxDiff, longitudinal and multilevel designs.

R, Python, SQL, SPSS, SAS, Stata, Mplus, BigQuery, Tableau, Power BI, Qualtrics, Excel.

Working together

This work comes in two forms: as a consultant through Significant Difference LLC, or as an employee. I am open to full-time and contract roles in research science, psychometrics, and people analytics. Details for hiring teams, including my CV.