EV-0049
A third of feedback interventions make performance worse
Kluger, A. N., & DeNisi, A. (1996). Psychological Bulletin, 119(2), 254-284 · Source
What this entitles us to say
A meta-analysis of 607 effect sizes from 23,663 observations finding that feedback interventions improved performance on average (d = .41) but that over a third of them made performance worse - and that effectiveness falls as feedback moves attention away from the task and toward the self.
What it found
607 effect sizes, 23,663 observations. Feedback interventions improved performance on average at d = .41 - and over one third of them decreased performance.
The explanatory finding: effectiveness falls as attention moves up the hierarchy, closer to the self and away from the task. Task-focused feedback enhances performance. Feedback directed at the self can reduce it, because attention shifts from doing the work to defending the person.
Why it matters here
Read the second finding against the five principles and something remarkable happens.
Principle two says the company's position is to point to the standard, not point out weaknesses. Principle three says shoulder to shoulder, not face to face - the Standard is the third thing in the room. Both were arrived at from conviction, not from research. And this meta-analysis says, in its own vocabulary, that feedback works when it points at the task and fails when it points at the self.
The principles and the research are the same finding in two languages. That is the strongest single piece of validation the program's design has received from outside itself, and it belongs in The Case for Character as well as in A16.
It also converts the principles from ethics into engineering. Privacy, the printed journal that goes home, the manager never seeing raw material, the aggregate-only summaries - these were justified as protections for the participant. This entry says they are also the conditions under which feedback works at all. A 360 that lands as a verdict on the person is in the third of interventions that make things worse.
Three specific design consequences:
1. The report must point at the standard, not the person. "Here is what Faithfulness looks like and here is where your ratings sit against it" - not "you are unreliable." 2. The 360 gap needs a task-shaped next step attached, or it is self-directed attention with nothing to do. 3. Never deliver a low score without a practice. Diagnosis alone is the failure mode this paper measured.
The limits
A 1996 meta-analysis, and the literature has continued since - but the core finding has not been overturned and remains the standard citation. The underlying studies are largely task-performance settings rather than leadership development. It concerns feedback interventions rather than 360 processes specifically, so the application to our design is by inference. That inference cuts toward caution, which is the safe direction.
What we do not say
"Our 360 avoids that problem." Nothing here has been tested on our instrument. The honest claim is that the design was built to respect the finding, not that it has been shown to.