When Feedback Fails

The most cited finding in feedback research is rarely quoted in management training: more than a third of feedback interventions degrade performance. Not mildly — measurably, consistently, across 607 effect sizes and 23,663 observations.[1] That result has survived every attempt to explain it away. It doesn't mean feedback is useless. It means the design of a feedback system is the thing that determines whether it helps or hurts — not the intention behind it.

I've been working through this literature alongside an internal effort to build a feedback culture in a mid-size French consulting division. The document we produced is a working tool, not a position paper. But the research it surfaced kept pulling me toward a harder question: why do most organisations treat feedback as if it were a free good — as if more of it, in any form, at any time, is obviously better?

The mechanism nobody talks about

Kluger & DeNisi's explanation of their paradox is the most useful piece of theory I encountered. They call it a displacement of attention. When feedback targets the task — "this analysis needed more quantitative grounding" — it improves performance. When it targets the person — "you're not rigorous" — it triggers identity defence. The cognitive resources that would otherwise go toward improving the work go instead toward managing the threat. Performance drops. Not because the message was wrong, but because the level at which it was pitched activated the wrong system.

Buckingham & Goodall extend this into three beliefs they call fallacies.[2] First: that evaluations reflect the qualities of the person being evaluated — they don't, they reflect the idiosyncrasies of the evaluator, the "idiosyncratic rater effect". Aggregating more raters doesn't solve this; it amplifies the error. Second: that critical feedback builds excellence — the neuroscience says otherwise, critical feedback activates the sympathetic nervous system and suppresses learning circuits. Third: that excellence is universal and transferable — what works for one manager can be actively counterproductive in another.

There is also a vulnerability differential that the literature underplays. Negative feedback is deleterious for individuals with low perceived power — junior staff, those without organisational leverage — and has no significant effect on those with high power.[3] The people who have the most to learn from honest feedback are also the most at risk from feedback that is clumsily framed. A system that doesn't account for this isn't neutral — it compounds existing inequalities.

The French context is not an excuse

Most of the feedback literature was written in Anglo-Saxon contexts and implicitly assumes them. France's Power Distance Index sits at 68, against roughly 40 for the United States and 20–35 for Nordic countries. A meta-analysis of over 190,000 employees across 32 countries shows that the effects of participative and upward feedback practices are empirically weaker in high-PDI cultures.[4] This is not a character flaw — it's a structural condition.

Philippe d'Iribarne's account of French working culture is useful here.[5] The model he describes is built on a hierarchy of ranks, each with its own obligations and dignities. In that frame, upward collective feedback isn't experienced as developmental input — it can be perceived as an attack on rank. The Anglo-Saxon literature takes the American model as the implicit norm and treats every deviation from it as a problem to fix. That's the wrong analytical move.

But the corollary matters: silence at work is not a French fatality. It's a rational calculation — risk of retaliation, perception that speaking up will change nothing, distrust of hierarchy.[6] A study across 33 countries and 8,222 respondents found that silence correlates more strongly with perceived organisational incentives than with national culture alone.[7] The silence is modifiable — but only if you change the incentives, not if you exhort people to be braver.

The literature distinguishes two types of silence that require different design responses. Defensive silence — triggered by perceived abusive management behaviour, a protective adaptation against predictable retaliation. And acquiescent silence — amplified by high PDI and the conviction that nothing will change anyway, dominant in bureaucratic or high-pressure environments.[8] Conflating them produces interventions that work for neither.

Four conditions that actually matter

The working document we built rests on four conditions the empirical literature converges on. Each is non-negotiable — remove any one, and the whole structure loses its claim to effectiveness.

Credible anonymity. A randomised field study of 38 managers showed that employees giving non-anonymous feedback attributed significantly higher scores.[9] The managers preferred the non-anonymous condition — it produced more positive feedback. More positive, less honest, therefore less useful. Anonymity doesn't make feedback kinder; it makes it real. In a high-PDI context, the architecture of anonymity — who sees what, at what level of aggregation, under what threshold — is not a technical detail. It's the whole system. The design we settled on evaluates organisational strata rather than individuals: junior staff rate the team leader level as a whole, not their specific team leader. This depersonalises the evaluation, reduces leniency bias, and produces macro-level information rather than interpersonal scores.

Feedback literacy. A meta-analysis of 24 longitudinal studies found that post-feedback improvement is generally weak without prior preparation — and is only measurable when four conditions exist: positive orientation toward feedback, perception that change is needed, positive reaction to the data, and belief that change is achievable.[10] Most managers don't arrive at those conditions spontaneously. Boud & Molloy define feedback literacy as the understanding, capacity, and disposition to make sense of feedback — and identify four trainable dimensions: appreciating feedback, forming judgements about one's own work, managing affect, and acting on the data.[11]

The dimension most neglected in standard management training is receiving. Almost all programmes teach managers to give feedback. The capacity to receive it — to sit with a score that's uncomfortable without becoming defensive or dismissive — is treated as personality, not skill. Research confirms it's teachable; it is nevertheless almost absent from standard curricula.[12]

Situational anchoring. Feedback that isn't anchored in specific, shared work situations slides toward personality assessment — which is both less accurate and more threatening. The question "does my manager communicate clearly?" invites a global character judgement. The question "in the last mission I completed, priorities were communicated clearly from the start" invites a factual recall. The second format is less coloured by affect, harder to distort, and produces information the manager can actually act on.[13] This matters especially in consulting structures where teams work in parallel on separate client portfolios — the shared situation is the mission, not the ongoing team dynamic.

Visible action cycle. A feedback system that collects data without producing visible organisational response is worse than no system. It confirms the acquiescent silence assumption: that nothing changes. A systematic review on post-survey processes establishes that the follow-through communication — "you said this; we did this" — is the primary determinant of participation in future cycles.[14] A broken promise is more damaging than no promise. The format we proposed: after each cycle, managers communicate what the collective data showed, what the organisation has decided to do in response, and the implementation timeline. Where a cause exceeds the manager's scope of action, name it explicitly and escalate — rather than making a commitment that can't be kept.

The trap the literature builds

One thing the research kept surfacing — and which deserves more direct naming than it usually gets — is the way positive psychology frameworks can function as social control mechanisms. Cabanas & Illouz document how corporate applications of positive psychology work by individualising the causes of discomfort: you lack resilience, you need to practise self-efficacy, your job crafting is underdeveloped.[15] The structure that generated the discomfort disappears from view. A review of 117 distinct critiques of positive psychology confirms that the constructs most commonly deployed in corporate wellbeing programmes — self-efficacy, grit, resilience — carry substantial unresolved academic criticism.[16]

Gill & Orgad are more pointed still: resilience, as valorised in contemporary management culture, is a neoliberal injunction that asks individuals to personally absorb the cost of their own recovery from conditions the organisation created.[17] A well-designed feedback system does the opposite — it identifies conditions, workload, information clarity, priority coherence, that the organisation can change. If the causes identified are structural, the response belongs at the directional level, not in a coaching conversation with the person who flagged them.

The sponsorship question turns on this. A feedback system that applies only to junior strata is experienced, correctly, as a control instrument. If the design is structurally sound — if the conditions above are genuinely met — then it should apply symmetrically, including to those who commissioned it.

The Gallup data for France is worth keeping in view throughout: 8% employee engagement in 2024, 36th out of 38 European countries.[18] Europe is already the least engaged region in the world. The decline is attributed primarily to deteriorating manager engagement. McKinsey's "Boss Factor" puts 70% of engagement variance on the direct manager.[19] Gallup adds that structured management training halves active disengagement.[18] The case for building this infrastructure isn't aspirational. The cost of the alternative is measurable.