The finding that changed how performance management should work

Kluger and DeNisi’s 1996 meta-analysis across 607 effect sizes and 23,663 observations produced a finding that the field largely ignored for decades: feedback interventions improved performance on average, but over one third of them decreased performance. Approximately one third improves performance, one third produces no effect, and one third actively makes performance worse. This distribution is not explained by whether the feedback was positive or negative, by the quality of the feedback, or by any variable that standard feedback intuition would predict.

The practical implication is stark. For performance reviews conducted without specific knowledge of when and how feedback works, the base rate of harm is substantial. A manager who delivers feedback without understanding the mechanisms governing its effects has roughly equal probability of helping, doing nothing, or making things worse. The widespread assumption that feedback is categorically beneficial is not supported by the most comprehensive evidence base available.

Why feedback harms: the locus of attention mechanism

Kluger and DeNisi’s Feedback Intervention Theory explains the pattern through a single critical variable: where the feedback directs the recipient’s attention. Their model identifies three levels. The task level focuses attention on the specific behaviour being evaluated and what can be improved. The self level focuses attention on what the feedback implies about the person’s identity, competence, and worth. The meta-task level focuses attention on strategies and learning processes.

Feedback that directs attention to the task level — specific, behavioural, focused on what was done and what can be done differently — tends to improve performance. Feedback that directs attention to the self level — evaluative of the person rather than the behaviour, connected to identity rather than to a specific act — tends to degrade performance, because it activates ego threat and the defensive psychological responses that ego threat produces.

The standard annual performance review almost universally activates the self level. It aggregates a year of behaviour into a holistic evaluation of the person, delivered in a formal context specifically designed to feel consequential. The formal context signals that the recipient’s standing in the organisation is being assessed. That signal is sufficient to shift attention from task processing to self-protection — and once attention is at the self level, the cognitive and motivational resources that would have supported developmental processing are redirected toward managing the identity threat.

A review that says “you are not a strategic thinker” activates the self level and produces ego threat. A review that says “in the Q3 presentation, addressing the competitive landscape earlier would have given the financial projections more context” activates the task level and produces developmental processing. The informational content can be identical. The locus of attention is different, and that difference determines the psychological response and the subsequent performance outcome.

Why the format itself is the problem

The standard performance review format activates every threat dimension of the SCARF model established earlier in this series. Status is threatened by formal evaluation against a standard. Certainty is threatened by the unpredictable nature of the assessment. Autonomy is threatened by being evaluated by an external authority. Relatedness is threatened by the transactional formality of the interaction. Fairness is threatened by the documented biases in performance ratings.

The SCARF threat profile of the standard review is not a function of the manager’s skill or intention. It is produced by the structural features of the format — the formal setting, the documented record, the connection to compensation, the year-end timing — independently of the quality of what is said. Designing better feedback content within the existing format addresses a secondary problem. The format itself is the primary mechanism generating the performance decrement.

Why accuracy is a separate and compounding problem

Even if the format were psychologically optimal, performance ratings carry a second, independent source of harm. Scullen, Mount and Goff’s research found that 61.5% of the variance in performance ratings reflects the idiosyncratic rating tendencies of the rater rather than the actual performance of the ratee. Recency bias, halo effects, and similarity bias contaminate the ratings systematically and reliably.

A person receiving a negative performance review in a system where the majority of rating variance reflects rater idiosyncrasy rather than their actual performance has a substantial probability of acting on feedback that does not accurately describe their behaviour. Changing behaviour that was not in fact the problem produces worse performance through misdirected effort. The accuracy problem and the ego threat problem are independent mechanisms that both degrade outcomes, operating simultaneously in the standard performance review format.

Positive feedback has its own trap

The counterintuitive finding extends to positive feedback. Research within the FIT framework found that positive feedback reduces subsequent effort when recipients have no new goal to pursue after receiving it. A performance review that tells someone they are excellent without defining the next level they should be reaching removes the performance discrepancy that was motivating their effort. The feedback closes the gap and eliminates the motivation the gap was generating.

The performance review that tells a high performer they are a high performer, without anchoring that evaluation to a higher standard or a new challenge, is a demotivational tool dressed as recognition.

What the Adobe natural experiment showed

Adobe eliminated its annual performance review in 2012, replacing it with frequent check-in conversations focused on real-time feedback and development. The result was a 30% reduction in voluntary attrition and reported improvements in engagement and manager-employee relationships. The check-in model maps directly onto FIT principles — frequent, specific, task-focused, future-oriented conversations rather than infrequent, holistic, identity-level evaluations.

The primary outcome Adobe observed was not a decline in accountability but the elimination of the psychological harm the formal annual evaluation had been producing. This is the practical validation of the Kluger and DeNisi finding at organisational scale.

The feedforward alternative

Kluger and DeNisi’s own research produced a practical structural alternative: feedforward. Rather than evaluating past performance, feedforward asks what the person would do differently in the future. The grammatical shift from past to future is also a psychological shift — from the self level, where past performance carries identity implications, to the task level, where future strategies carry no self-threat. The cognitive resources that would have been consumed managing ego threat are available for developmental processing. Feedforward consistently outperforms traditional feedback in experimental settings because it bypasses the ego threat mechanism entirely.

A book worth reading alongside this

Thanks for the Feedback by Douglas Stone and Sheila Heen is the most practically grounded available treatment of why feedback fails and what can be done about it. Stone and Heen’s three-trigger framework — truth triggers, which fire when the feedback feels factually wrong; relationship triggers, which fire when the relationship with the person giving feedback contaminates the reception of it; and identity triggers, which fire when the feedback threatens the recipient’s sense of who they are — maps directly onto the FIT self-level attention mechanism and the ego threat research this article draws on. Their research on how different people’s identity stability determines their vulnerability to the identity trigger is particularly useful for any entrepreneur managing a team with varying levels of psychological resilience. The book is rigorous without being inaccessible, and it provides the most direct available translation of the Kluger and DeNisi framework into management behaviour that can be changed on Monday morning.

Have questions about this article?

If any part of this article raised questions you want to explore further, courbot.co is built for exactly that. It is courben.co’s AI assistant, designed around the psychology of entrepreneurship. Ask it anything from this article.

This article is for educational and informational purposes only.

This article is for educational and informational purposes only. Sources: Kluger, A.N. & DeNisi, A. (1996), Psychological Bulletin, 119(2), 254–284. Baumeister, R.F., Heatherton, T.F. & Tice, D.M. (1993), Journal of Personality and Social Psychology, 64(1). Scullen, S.E., Mount, M.K. & Goff, M. (2000), Journal of Applied Psychology, 85(6), 956–970. Rock, D. (2008), NeuroLeadership Journal, 1(1).