peopleanalyst

Magazine · long-form

principal-issues.

Long-form on one idea worth defending: behavioral science — the discipline of measuring people and how they change — is the missing methodology for AI, not its casualty. Each essay takes one place the AI field is reinventing something psychometrics or diffusion science already settled, and shows the older answer. Measurement-first, source-anchored, claims defended rather than asserted — and an experiment in bi-directional adaptive learning.

Currently a small set of pieces, growing as the program does. Borrowing infrastructure (and editorial discipline) from Vela's magazine, oriented to a different topic domain.

Read as a set · 8 principles

The Principles — how we think about measuring people, and AI

The handful of essays that articulate our philosophy and how we build — gathered so they don't crowd the magazine. Each also stands alone as a piece of method you can use whether or not you ever work with us.

Read the principles →

Read as a set · 7 pieces · field craft

The ways a dashboard lies — and how to read it honestly

The recurring traps that turn a confident chart into a wrong decision — causation, composition, small samples, survivorship, multiple comparisons. Pure method, free to use. (These also appear in the feed below.)

Read field craft →

Read as a set · 4 pieces · what the tools miss

What the tools miss — performance, one setting at a time

A cited essay per context — engineering, the support floor, the hospital, the school — each leading with what performance actually means there, and what generic, off-the-shelf tools miss. (These also appear in the feed below.)

Read what the tools miss →

Methodology · regression to the mean

people analytics · organization measurement & data science · AI–human interaction

June 28, 2026

The Turnaround That Wasn't

Pull the lowest-scoring teams, run a program, and watch their scores jump — the biggest gains in the company. The jump is mostly regression to the mean: a group selected for being at the bottom climbs back toward the middle whether or not you did anything. Without a control, the slide can't tell.

Twenty bottom-decile managers go through a development program and post the largest year-over-year engagement gains anywhere; the offsite calls it a win and scales it. But they were chosen for scoring at the bottom, and a group chosen for an extreme drifts back toward the mean on re-measure regardless — because part of any extreme score is transient noise that doesn't repeat. Galton named the effect in 1886; Kahneman's flight instructors misread it as praise hurting and rebuke helping. The honest design builds the one comparison that sees through it: randomize within the bottom decile, or run difference-in-differences against equally-low teams that got nothing — the regression hits both, so the gap left is your effect. It costs you the clean two-bar slide and answers the only question that matters: did we cause the improvement, or schedule a measurement at the bottom of a bounce? Getting easier to miss as targeting automates — a model flags the high-risk, the scores subside, and the dashboard takes the credit. Before you scale the turnaround: did we pick these units for being extreme on the same measure we now use to score them, and was there a group just as bad that got nothing?

Read · by Mike West