The People Analyst Guide to Nine Lies About Work. Format: what the book argues → what the research actually says → how you run it → the analysis you can run → the AI-era turn → what to do Monday. No reproduction of the book's text; the substance is ours. Research anchors verified on read.
What the book argues
The fifth lie is "people need feedback." The book's counter is that what people actually need is attention — specifically attention to what they do well — and that the modern feedback apparatus (more reviews, more 360s, the radical-candor reflex to tell people what's wrong) is largely counterproductive. The provocation lands because it inverts an article of faith: that the helpful, generous thing is to point out the gap.
What the research actually says
This is another lie that is closer to true than it sounds, and the evidence predates the book by two decades. The foundational result is Kluger & DeNisi's (1996) feedback-intervention meta-analysis: feedback's effect on performance is highly variable, and a substantial share of feedback interventions — on the order of a third — reduced performance. The mechanism is the part to hold onto: feedback helps when it directs attention to the task, and hurts when it directs attention to the self — the ego, the comparison, the rating. Evaluative, identity-directed feedback (which is most of what performance management delivers) is exactly the kind that backfires.
The honest caveat the Guide must keep: this is not "never give feedback." Specific, timely, task-focused feedback — the cue that tells you what to adjust on the next rep — is among the most reliable performance levers there is. The lie overstates by erasing that. The truth is narrower and more useful: most feedback is delivered in the form that backfires, and organizations measure its volume instead of its effect. We mandate quarterly feedback and count completion; we almost never check whether performance moved afterward.
How you run it
Measure feedback by its consequence, not its frequency. Three moves: classify feedback as task- vs. self-directed (the variable that decides whether it helps); measure the effect — did the metric move after the intervention, against a comparison that didn't get it; and measure attention to what works (recognition, strengths use) as its own signal rather than a soft add-on. Stop rewarding managers for delivering feedback and start asking whether the feedback did anything.
The analysis you can run
This is a feedback / recognition-effect analysis — program-evaluation with survey-orchestrator —
that treats a feedback or recognition practice as an intervention and asks the only question that
matters: did the outcome change, for whom, and in which direction? It separates the task-directed cues
that help from the ego-directed evaluation that backfires, and it pairs with the rater-reliability work
(Lie 6) — because a feedback rating you can't trust is worse than none. (Ties to our
feedback-backfires proof point.)
The AI-era turn
AI now writes feedback at scale — drafted reviews, auto-generated coaching, sentiment nudges. Run through this lie, scale is the danger: it industrializes exactly the ego-directed, evaluative feedback that Kluger & DeNisi flagged, in fluent prose that feels helpful. The fix isn't to ban it; it's to measure its effect before scaling it, and to bias the machine toward task-focused, forward-looking cues rather than verdicts on the person.
What to do Monday
- Stop counting feedback volume. Completion rates are not impact.
- For one feedback or recognition practice, measure whether performance moved against a comparison — treat it as an experiment.
- Shift the register from evaluative to task-focused ("here's the adjustment for next time", not "here's your rating"); ego-directed feedback is the kind that backfires.
- Before scaling any AI-written feedback, test its effect like any other intervention.