Feedback: what research promises, and what it refuses to promise
Few managerial practices enjoy consensus as broad as feedback. Few hold up as poorly under empirical scrutiny. The useful question is not 'should we give feedback?' but 'under what conditions does a response produce learning rather than deterioration?'. The available evidence allows a partial answer, and demands caution.
The finding rarely quoted
The meta-analysis by Kluger and DeNisi (1996, Psychological Bulletin) remains the reference: across several hundred experimental effects, feedback interventions improve performance on average, yet in a substantial share of cases — the authors report roughly one third — they degrade it. This does not invalidate the practice; it invalidates the idea that a response is beneficial by nature. Thirty years on, no robust refutation of that result has been established.
The proposed explanation: where attention lands
On the theory advanced by the same authors, the effect depends on the level toward which the response directs attention. When it bears on the task and on what can be adjusted, it supports learning. When it bears on the self — the person's worth, their comparison with others — it diverts cognitive resources toward protecting self-esteem, at the task's expense. This is a supported theoretical account, not a law: the experimental designs often involve simple tasks, and transposing them to real managerial work remains an inference.
What this invalidates in common practice
The so-called sandwich method, wrapping criticism between two compliments, blurs exactly that level of attention: the recipient retains the overall verdict, not the requested adjustment. The annual review, aggregating twelve months of facts into one summary judgement, produces the same shift from content to status. Both observations are consistent with the described mechanism; they do not prove that any other form works.
What the literature on reception adds
Yip and Fisher (2022, Academy of Management Annals) move attention from the giver to the receiver: the value of a response depends on readiness to receive it, which depends on the relationship, perceived intent and timing. An uncomfortable implication: feedback quality cannot be decreed through a catalogue of phrasings. It is built over the life of a working relationship — which is why an isolated session on the art of feedback rarely yields lasting change without managerial support. (our executive and employee training programmes)
Three defensible conditions, and their limits
What current knowledge cautiously supports: a response carries further when it targets an observable behaviour rather than a trait; when it is close to the event rather than deferred; when it opens onto a specific action rather than an evaluation. None of the three is sufficient, and their effect is not uniformly established across contexts. They form a reasonable working frame, not a guarantee.
Implication for Swiss executive teams
Before investing in new appraisal tooling, it is cheaper to check one thing: do the responses given inside the organisation bear on identifiable behaviours, or on people? The answer comes from reading a sample of review write-ups, with no instrumentation required. It is also the starting point we use in our engagements in Geneva, ahead of any training. To go further: explore the The Art of Feedback training in Geneva, or browse our executive and employee training programmes in Switzerland.
In pictures: The Art of Feedback in Geneva
- the art of feedback
- Geneva
- research
- training geneva