topic · exploring

Feedback Interventions That Backfire

What we actually think: the popular management axiom that feedback is an unalloyed good is contradicted by the largest meta-analysis ever run on the question. Kluger & DeNisi (1996) — 131 usable papers, 607 effect sizes, 12,652 participants, 23,663 observations, all re-confirmed against the published PDF this pass — found an average benefit (d = 0.41) wrapped around a negative tail: over 38% of effects were negative, 32% even after trimming outliers and dependence violations, with observed variance (0.97) an order of magnitude above what sampling error explains (0.09). The right reading is not "feedback is bad" but "feedback is a high-variance intervention whose sign is governed by where it points attention."

Feedback Intervention Theory is the load-bearing explanation, and its moderator table (Table 2, verified cell by cell) is the calibration: cues that keep attention on the task help — correct solution d = 0.43 vs 0.25, velocity d = 0.55 vs 0.28, accompanying goal setting d = 0.51 vs 0.30 — while cues that pull attention to the self hurt — praise d = 0.09 vs 0.34, top-quartile threat to self-esteem d = 0.08 vs 0.47, feedback designed to discourage d = −0.14. Task complexity moderates too (0.03 top quartile vs 0.55 bottom). The education literature converges independently: Wisniewski, Zierer & Hattie (2020; primary-checked) put high-information feedback at d = 0.99 against 0.24 for bare reinforcement/punishment, with 17% of effects negative. We hold the moderator claims at "moderate" because they are meta-analytic codings, not randomized head-to-head tests — the report's own caveat, which we keep visible.

The bite for leadership practice: the standard rituals cluster on the self side of the line. Multisource (360) feedback yields corrected improvement effects of d = 0.05–0.15 with CIs including zero for all sources but supervisors (Smither et al. 2005, primary-checked, including the authors' "unrealistic for practitioners" conclusion); the ratings feeding those systems are 53–62% rater idiosyncrasy versus 21–25% actual ratee performance (Scullen et al. 2000, primary-checked); the feedback sandwich rests on two conflicting experiments (held "contested"); and the claim that rituals aim at the self is honestly a theory-based audit, not a measurement — we hold it at "suggestive." Both ritual-level claims sit at "moderate" rather than the report's "strong" because each rests on a single supporting source.

The counter-evidence is the thesis's other half, not its embarrassment. Goals need feedback (Locke & Latham 2002, primary-checked: "the combination of goals plus feedback is more effective than goals alone"; Neubert 1998: incremental d = 0.63), and 140 randomized trials of audit-and-feedback show a median +4.3-point compliance gain with an all-positive IQR — moderators that read like a FIT checklist (Ivers et al. 2012, primary-checked; the 2025 update with 292 studies stays consistent). Wise-feedback experiments show candid criticism lands when high standards and belief in the recipient are explicit (Yeager et al. 2014, primary-checked: 71% vs 17% adjusted revision rates, from cells of ~11 at p = .045 — trust the sign, not the size). The synthesis: feedback aimed at the work is among the most reliable interventions in applied psychology; feedback aimed at the person is a coin flip weighted slightly toward harm.

Precision notes from this ingestion, so downstream prose does not overclaim: the abstract says "over 1/3," the results section "over 38%" — of effects in included studies, not of all feedback ever given (natural feedback-seeking was out of scope); a distribution with mean 0.41 and SD near 1 mechanically puts about a third of its mass below zero, so the interesting finding is that FIT's moderators predict which effects land in the tail; and the Ivers +4.3% median comes from the 82 low-risk-of-bias dichotomous-outcome comparisons (49 studies), not all 140 trials — a detail the seed report glossed. The report's attention-allocation equations (E[d] = μ + βT·T − βS·S and the delivery decision rule) are managerial syntheses the report itself marks as theoretical; we did not encode them as claims, and any figure built on them must be labeled as a model calibrated to Table 2, not as fitted data.

strongproposedclm.feedback-interventions-backfire.negative-tail

In the largest meta-analysis of feedback interventions (607 effect sizes from 131 papers, 12,652 participants, 23,663 observations), feedback improved performance on average (d = 0.41) yet over 38% of effect sizes were negative — a share that survives outlier trimming and exclusion of dependence violations (32% of 470 trimmed effects) and cannot be attributed to sampling error (weighted variance 0.97 vs expected 0.09).

supportsprimary-checkedAbstract, p. 254
“A meta-analysis (607 effect sizes; 23,663 observations) suggests that FIs improved performance on average (d = .41) but that over 1/3 of the FIs decreased performance. This finding cannot be explained by sampling error, feedback sign, or existing theories.”
weighted mean Cohen's d of feedback interventions on performance: 0.41 (n = 607 effect sizes; 12,652 participants; 23,663 observations)
Kluger, A., DeNisi, A. (1996). The Effects of Feedback Interventions on Performance: A Historical Review, a Meta-Analysis, and a Preliminary Feedback Intervention Theory. Psychological Bulletin, 119(2), 254-284. doi:10.1037/0033-2909.119.2.254
supportsprimary-checkedp. 258
“The weighted mean (weighted by sample size) of this distribution is 0.41, suggesting that, on average, FI has a moderate positive effect on performance. However, over 38% of the effects were negative (see Figure 1). The weighted variance of this distribution is 0.97, whereas the estimate of the sampling error variance is only 0.09.”
share of negative effects; weighted variance vs expected sampling-error variance: >38% negative; variance 0.97 vs 0.09 expected
Kluger, A., DeNisi, A. (1996). The Effects of Feedback Interventions on Performance: A Historical Review, a Meta-Analysis, and a Preliminary Feedback Intervention Theory. Psychological Bulletin, 119(2), 254-284. doi:10.1037/0033-2909.119.2.254
supportsprimary-checkedp. 273
“Overall, 470 effect sizes survived all exclusions (d′″) and contained enough information to be rated on at least one moderator. Of these effects, 32% were negative. The average FI effect was .38 with a variance of .45 (drastically reduced because of the trimming of the outliers), whereas the expected variance was .09.”
trimmed dataset (outliers, Mikulincer effects, and quasi-d time-series removed): mean d = 0.38; 32% negative; variance 0.45 vs 0.09 expected (n = 470 effect sizes)
Kluger, A., DeNisi, A. (1996). The Effects of Feedback Interventions on Performance: A Historical Review, a Meta-Analysis, and a Preliminary Feedback Intervention Theory. Psychological Bulletin, 119(2), 254-284. doi:10.1037/0033-2909.119.2.254
contextualizesprimary-checkedTable 2, p. 273 (verbatim table values)
“Task complexity (P3) — Top quartile: K = 107, d = .03; Bottom quartile: K = 114, d = .55.”
mean d by task complexity quartile (after all exclusions): 0.03 (most complex) vs 0.55 (simplest) (n = K = 107 vs 114 effect sizes)
Kluger, A., DeNisi, A. (1996). The Effects of Feedback Interventions on Performance: A Historical Review, a Meta-Analysis, and a Preliminary Feedback Intervention Theory. Psychological Bulletin, 119(2), 254-284. doi:10.1037/0033-2909.119.2.254
supportsprimary-checkedp. 155, Literature Review — NOTE: the aggregate includes Kluger & DeNisi itself; only the Bangert-Drowns et al. 1991 member is independent, and the unit is studies, not effects
“In fact, about one third of the total studies reviewed in two landmark meta-analyses (i.e., Bangert-Drowns et al., 1991; Kluger & DeNisi, 1996) demonstrate negative effects of feedback on learning.”
Shute, V. (2008). Focus on Formative Feedback. Review of Educational Research, 78(1), 153-189. doi:10.3102/0034654307313795
contextualizesprimary-checkedFigure 1 caption, p. 258 — the paper prints the raw histogram itself; its x-axis runs to d = 12.5 via an axis break, with the visible bulk between roughly −2 and +3
“Figure 1. Distribution (histogram) of 607 effects (ds) of feedback intervention on performance.”
Kluger, A., DeNisi, A. (1996). The Effects of Feedback Interventions on Performance: A Historical Review, a Meta-Analysis, and a Preliminary Feedback Intervention Theory. Psychological Bulletin, 119(2), 254-284. doi:10.1037/0033-2909.119.2.254
contextualizesprimary-checkedp. 273, Results
“In addition, three moderators almost reached our significance criteria of .01 (i.e., ps < .05): Computerized FI yielded stronger FI effects (consistent with P2); FIs on complex tasks yielded weaker effects (P3); and FIs were more effective with a goal-setting intervention (P4). Yet, these effects should be treated with extra caution because of the reasons that led us to set alpha at .01 above.”
Kluger, A., DeNisi, A. (1996). The Effects of Feedback Interventions on Performance: A Historical Review, a Meta-Analysis, and a Preliminary Feedback Intervention Theory. Psychological Bulletin, 119(2), 254-284. doi:10.1037/0033-2909.119.2.254
contextualizesprimary-checkedp. 275, Moderator Analyses: Major Conclusions
“Task complexity had relatively low interjudge reliability (.70), reflecting perhaps the difficulty in conceptualizing task complexity (and other task dimensions) and therefore suggesting that the effect that we observed is an underestimate. Indeed, when we investigated the meaning of the weak correlational effect of task complexity with differences in mean FI effect between the extreme quartiles of task complexity (Table 2), a large effect of task complexity appeared. (Of course, this effect appears large because we looked at the extreme quartiles, yet it helps to demonstrate the implication of the weak correlation.)”
Kluger, A., DeNisi, A. (1996). The Effects of Feedback Interventions on Performance: A Historical Review, a Meta-Analysis, and a Preliminary Feedback Intervention Theory. Psychological Bulletin, 119(2), 254-284. doi:10.1037/0033-2909.119.2.254
contextualizesprimary-checkedp. 274
“Table 2 also suggests that even within each level of the moderators, there is a large portion of unexplained variance of FI effects.”
Kluger, A., DeNisi, A. (1996). The Effects of Feedback Interventions on Performance: A Historical Review, a Meta-Analysis, and a Preliminary Feedback Intervention Theory. Psychological Bulletin, 119(2), 254-284. doi:10.1037/0033-2909.119.2.254
contextualizesreport-derivedSeed report, Key sources table
“Feedback is "a double-edged sword"; praise-type FIs d ≈ 0.09 vs 0.34 without”
Kluger, A., DeNisi, A. (1998). Feedback Interventions: Toward the Understanding of a Double-Edged Sword. Current Directions in Psychological Science, 7(3), 67-72. doi:10.1111/1467-8721.ep10772989

Counter-evidence searched: Searched 2026-07-07 for rebuttals of the negative-tail figure; found no published dispute of the number itself. Two scoping caveats are incorporated instead of hidden: (1) the base is effects in the included studies — natural feedback-seeking and intrinsic feedback were out of scope — so 'one third of all feedback' overstates it; (2) any distribution with mean 0.41 and SD near 1 mechanically puts roughly a third of its mass below zero, so the scientifically interesting finding is that FIT's moderators predict which effects land in the tail, not the tail's existence. Corroboration caveat (2026-07-09): Shute (2008)'s one-third figure explicitly aggregates two meta-analyses, one of which is Kluger & DeNisi itself (the other being Bangert-Drowns et al. 1991, education-specific), and counts studies, not effects — so its support runs only through the aggregate's independent member (Bangert-Drowns et al. 1991, education-specific); the Kluger & DeNisi share is echo, not a second discovery — hence the essay presents it as corroboration with the circularity named. Predictability caveat: even within each moderator level the paper reports a large portion of unexplained variance (p. 274), so the moderators aim the distribution without pinning individual interventions.

moderateproposedclm.feedback-interventions-backfire.self-cues-attenuate

Feedback cues that direct attention to the self rather than the task — praise, discouragement, threats to self-esteem — attenuate or reverse feedback's average benefit: meta-analytic moderator cells show d = 0.09 with praise vs 0.34 without, d = 0.08 in the top quartile of self-esteem threat vs 0.47 in the bottom, and d = −0.14 for feedback designed to discourage.

supportsprimary-checkedp. 267
“Proposition 1: FI effects on performance are attenuated by cues that direct attention to meta-task processes (P1). Such cues include normative FIs, person-mediated versus computer-mediated FIs, FIs designed either to discourage or praise the person, and any cue that may be perceived as a threat to the self.”
Kluger, A., DeNisi, A. (1996). The Effects of Feedback Interventions on Performance: A Historical Review, a Meta-Analysis, and a Preliminary Feedback Intervention Theory. Psychological Bulletin, 119(2), 254-284. doi:10.1037/0033-2909.119.2.254
supportsprimary-checkedTable 2, p. 273 (verbatim table values)
“Praise (P1) — Yes: K = 80, d = .09; No: K = 358, d = .34. Threat to self-esteem (P1) — Top quartile: K = 102, d = .08; Bottom quartile: K = 170, d = .47. Discouraging FI (P1) — Yes: K = 49, d = −.14; No: K = 388, d = .33.”
mean d by self-cue moderator (after all exclusions): praise 0.09 vs 0.34; self-esteem threat 0.08 (top quartile) vs 0.47 (bottom); discouraging −0.14 vs 0.33 (n = K = 80/358 (praise), 102/170 (threat), 49/388 (discouraging) effect sizes)
Kluger, A., DeNisi, A. (1996). The Effects of Feedback Interventions on Performance: A Historical Review, a Meta-Analysis, and a Preliminary Feedback Intervention Theory. Psychological Bulletin, 119(2), 254-284. doi:10.1037/0033-2909.119.2.254
supportsprimary-checkedp. 275, Moderator Analyses: Major Conclusions
“Specifically, both praise and FI designed to discourage were postulated to increase attention to meta-task processes and were found to attenuate FI effects.”
Kluger, A., DeNisi, A. (1996). The Effects of Feedback Interventions on Performance: A Historical Review, a Meta-Analysis, and a Preliminary Feedback Intervention Theory. Psychological Bulletin, 119(2), 254-284. doi:10.1037/0033-2909.119.2.254
supportsprimary-checkedp. 275, Moderator Analyses: Major Conclusions
“Furthermore, both the attenuating effect of praise and the nonsignificant effect of an FI sign (which is discussed later in this section) are not easily predicted by most FI-related theories.”
Kluger, A., DeNisi, A. (1996). The Effects of Feedback Interventions on Performance: A Historical Review, a Meta-Analysis, and a Preliminary Feedback Intervention Theory. Psychological Bulletin, 119(2), 254-284. doi:10.1037/0033-2909.119.2.254
supportsreport-derivedSeed report, §Evidence landscape
“children praised for intelligence dropped 0.92 problems after failure while children praised for effort gained 1.21”
post-failure change in problems solved (intelligence vs effort praise): −0.92 vs +1.21; F(2,120) = 17.62 (per seed report) (n = 128 fifth graders (Study 1))
Mueller, C., Dweck, C. (1998). Praise for Intelligence Can Undermine Children's Motivation and Performance. Journal of Personality and Social Psychology, 75(1), 33-52. doi:10.1037/0022-3514.75.1.33
supportsreport-derivedSeed report, §Evidence landscape
“fifth-graders given task-involving comments improved while those given ego-involving grades (or grades plus comments) did not”
classroom performance by feedback condition: comments improved; grades and grades+comments did not (per seed report) (n = 132 pupils analyzed, 12 classes)
Butler, R. (1988). Enhancing and Undermining Intrinsic Motivation: The Effects of Task-Involving and Ego-Involving Evaluation on Interest and Performance. British Journal of Educational Psychology, 58(1), 1-14. doi:10.1111/j.2044-8279.1988.tb00874.x
supportsreport-derivedSeed report, Key sources table
“four feedback levels; feedback about the self as a person is least effective; praise ≈ 0.12–0.14”
Hattie, J., Timperley, H. (2007). The Power of Feedback. Review of Educational Research, 77(1), 81-112. doi:10.3102/003465430298487
supportsreport-derivedSeed report, §Evidence landscape
“Recipients of unfavorable 360 feedback reacted with anger and discouragement and judged the feedback less accurate and less useful — "results question widely held assumptions... that negative and discrepant feedback motivates positive change"”
Brett, J., Atwater, L. (2001). 360-Degree Feedback: Accuracy, Reactions, and Perceptions of Usefulness. Journal of Applied Psychology, 86(5), 930-942. doi:10.1037/0021-9010.86.5.930

Counter-evidence not yet searched.

moderateproposedclm.feedback-interventions-backfire.task-cues-augment

Feedback features that keep attention on the task augment its effect: meta-analytic moderator cells show d = 0.43 with the correct solution vs 0.25 without, d = 0.55 for velocity feedback (progress against one's own past performance) vs 0.28, d = 0.51 with accompanying goal setting vs 0.30, and in education high-information feedback averages d = 0.99 vs 0.24 for bare reinforcement or punishment.

supportsprimary-checkedp. 268
“Proposition 2: FI effects on performance are augmented by (a) cues that direct attention to task-motivation processes and (b) cues that direct attention to task-learning processes coupled with information regarding erroneous hypotheses (P2).”
Kluger, A., DeNisi, A. (1996). The Effects of Feedback Interventions on Performance: A Historical Review, a Meta-Analysis, and a Preliminary Feedback Intervention Theory. Psychological Bulletin, 119(2), 254-284. doi:10.1037/0033-2909.119.2.254
supportsprimary-checkedTable 2, p. 273 (verbatim table values)
“Correct solution (P2) — Yes: K = 114, d = .43; No: K = 197, d = .25. Velocity (P2) — Yes: K = 50, d = .55; No: K = 380, d = .28. Goal setting (P4) — Yes: K = 37, d = .51; No: K = 373, d = .30.”
mean d by task-cue moderator (after all exclusions): correct solution 0.43 vs 0.25; velocity 0.55 vs 0.28; goal setting 0.51 vs 0.30 (n = K = 114/197 (solution), 50/380 (velocity), 37/373 (goals) effect sizes)
Kluger, A., DeNisi, A. (1996). The Effects of Feedback Interventions on Performance: A Historical Review, a Meta-Analysis, and a Preliminary Feedback Intervention Theory. Psychological Bulletin, 119(2), 254-284. doi:10.1037/0033-2909.119.2.254
contextualizesprimary-checkedp. 278, Implications
“Specifically, an FI provided for a familiar task, containing cues that support learning, attracting attention to feedback-standard discrepancies at the task level (velocity FI and goal setting), and is void of cues to the meta-task level (e.g., cues that direct attention to the self) is likely to yield impressive gains in performance, possibly exceeding 1 SD.”
Kluger, A., DeNisi, A. (1996). The Effects of Feedback Interventions on Performance: A Historical Review, a Meta-Analysis, and a Preliminary Feedback Intervention Theory. Psychological Bulletin, 119(2), 254-284. doi:10.1037/0033-2909.119.2.254
contextualizesprimary-checkedp. 273, Results
“In addition, three moderators almost reached our significance criteria of .01 (i.e., ps < .05): Computerized FI yielded stronger FI effects (consistent with P2); FIs on complex tasks yielded weaker effects (P3); and FIs were more effective with a goal-setting intervention (P4). Yet, these effects should be treated with extra caution because of the reasons that led us to set alpha at .01 above.”
Kluger, A., DeNisi, A. (1996). The Effects of Feedback Interventions on Performance: A Historical Review, a Meta-Analysis, and a Preliminary Feedback Intervention Theory. Psychological Bulletin, 119(2), 254-284. doi:10.1037/0033-2909.119.2.254
supportsprimary-checkedAbstract
“Overall results based on a random-effects model indicate a medium effect (d = 0.48) of feedback on student learning, but the significant heterogeneity in the data shows that feedback cannot be understood as a single consistent form of treatment.”
overall d of educational feedback (post-outlier random-effects estimate): 0.48 (n = 994 effects, 435 studies, N > 61,000)
Wisniewski, B., Zierer, K., Hattie, J. (2020). The Power of Feedback Revisited: A Meta-Analysis of Educational Feedback Research. Frontiers in Psychology, 10, 3087. doi:10.3389/fpsyg.2019.03087
supportsprimary-checkedTable 3 (verbatim row values)
“High-information feedback 42 0.99 [0.82 – 1.15] ... Reinforcement or punishment ... 0.24”
d by feedback information content: high-information 0.99 vs reinforcement/punishment 0.24 (n = 42 high-information effects (of 994))
Wisniewski, B., Zierer, K., Hattie, J. (2020). The Power of Feedback Revisited: A Meta-Analysis of Educational Feedback Research. Frontiers in Psychology, 10, 3087. doi:10.3389/fpsyg.2019.03087
supportsreport-derivedSeed report, §Evidence landscape
“the incremental value of adding feedback to goals is d = 0.63 across 16 effect sizes”
incremental d of feedback added to goal setting: 0.63 (per seed report) (n = 16 effect sizes)
Neubert, M. (1998). The Value of Feedback and Goal Setting over Goal Setting Alone and Potential Moderators of This Effect: A Meta-Analysis. Human Performance, 11(4), 321-335. doi:10.1207/s15327043hup1104_2
contextualizesreport-derivedSeed report, Key sources table
“first guideline: "Focus feedback on the task, not the learner"”
Shute, V. (2008). Focus on Formative Feedback. Review of Educational Research, 78(1), 153-189. doi:10.3102/0034654307313795

Counter-evidence not yet searched.

moderateproposedclm.feedback-interventions-backfire.multisource-small

Longitudinal multisource (360-degree) feedback programs produce at best small improvement in subsequent ratings — corrected mean d = 0.15 (direct reports), 0.05 (peers), 0.15 (supervisors), and −0.04 (self) across 24 longitudinal studies — with 95% confidence intervals including zero for every source except supervisors, and the outcome is rating change, not objective performance.

supportsprimary-checkedResults, pp. 40–41
“For direct report feedback, 19 of the 21 effect sizes were positive but the corrected mean effect size was only .15. For peer feedback, 6 of the 7 effect sizes were positive but the corrected mean effect size was only .05. For supervisor feedback, 8 of the 10 effect sizes were positive but the corrected mean effect size was only .15. For self-ratings, 6 of the 11 effect sizes were positive; the corrected mean effect size was −.04. For each rater source except supervisors, the 95% confidence interval included zero.”
corrected mean d of rating improvement by source: 0.15 (direct reports), 0.05 (peers), 0.15 (supervisors), −0.04 (self) (n = 24 longitudinal studies; N = 7,705 / 5,331 / 5,358 / 3,684 per source)
Smither, J., London, M., Reilly, R. (2005). Does Performance Improve Following Multisource Feedback? A Theoretical Model, Meta-Analysis, and Review of Empirical Findings. Personnel Psychology, 58(1), 33-66. doi:10.1111/j.1744-6570.2005.514_1.x
supportsprimary-checkedConclusion, p. 60
“The results of this meta-analysis and evidence related to the theoretical model indicate that it is unrealistic for practitioners to expect large across-the-board performance improvement after people receive multisource feedback. Instead, it appears that some feedback recipients will be more likely to improve than others.”
Smither, J., London, M., Reilly, R. (2005). Does Performance Improve Following Multisource Feedback? A Theoretical Model, Meta-Analysis, and Review of Empirical Findings. Personnel Psychology, 58(1), 33-66. doi:10.1111/j.1744-6570.2005.514_1.x
contextualizesreport-derivedSeed report, Key sources table
“Managers who met with direct reports to discuss feedback improved more; overall improvement moderate (d ≈ 0.51 per Heslin & Latham 2004)”
Walker, A., Smither, J. (1999). A Five-Year Study of Upward Feedback: What Managers Do with Their Results Matters. Personnel Psychology, 52(2), 393-423. doi:10.1111/j.1744-6570.1999.tb00166.x

Counter-evidence not yet searched.

moderateproposedclm.feedback-interventions-backfire.ratings-measure-rater

In large multirater datasets, idiosyncratic rater effects account for 62% and 53% of the variance in job performance ratings, while actual ratee performance (general plus dimensional) accounts for only 21% and 25% — the ratings that feed appraisal rituals mostly measure the rater, not the ratee.

supportsprimary-checkedAbstract (via PubMed 11125659)
“Idiosyncratic rater effects (62% and 53%) accounted for over half of the rating variance in both data sets.”
share of performance-rating variance attributable to idiosyncratic rater effects: 62% and 53% across two datasets (n = 2,350 and 2,142 managers, 7 raters each)
Scullen, S., Mount, M., Goff, M. (2000). Understanding the Latent Structure of Job Performance Ratings. Journal of Applied Psychology, 85(6), 956-970. doi:10.1037/0021-9010.85.6.956
supportsprimary-checkedAbstract (via PubMed 11125659)
“The combined effects of general and dimensional ratee performance (21% and 25%) were less than half the size of the idiosyncratic rater effects.”
share of rating variance attributable to ratee performance (general + dimensional): 21% and 25% across two datasets
Scullen, S., Mount, M., Goff, M. (2000). Understanding the Latent Structure of Job Performance Ratings. Journal of Applied Psychology, 85(6), 956-970. doi:10.1037/0021-9010.85.6.956

Counter-evidence not yet searched.

contestedproposedclm.feedback-interventions-backfire.sandwich-untested

The feedback sandwich, despite its ubiquity in management practice, rests on essentially two direct experiments that point in opposite directions: one found it changes perceptions but not performance, the other weakly favored it, possibly from mere added positivity — an evidence vacuum, not a validated technique.

supportsreport-derivedSeed report, Key sources table (quoting Parkes et al. 2013)
“Students think feedback sandwiches positively impact subsequent performance when there is no evidence that they do”
sandwich effect on perceptions vs performance: perceptions improved; performance did not (per seed report) (n = N = 20 and N = 350 (two studies))
Parkes, J., Abercrombie, S., McCarty, T. (2013). Feedback Sandwiches Affect Perceptions but Not Performance. Advances in Health Sciences Education, 18(3), 397-407. doi:10.1007/s10459-012-9377-9
contradictsreport-derivedSeed report, Key sources table
“Sandwich group spent more time preparing and solved more problems — "only partial evidence for the effectiveness of sandwich feedback," possibly from mere positivity”
sandwich vs non-sandwich feedback on task performance: sandwich group performed better (per seed report) (n = N = 91, three arms)
Prochazka, J., Ovcari, M., Durinik, M. (2020). Sandwich Feedback: The Empirical Evidence of Its Effectiveness. Learning and Motivation, 72, 101649. link

Counter-evidence not yet searched.

suggestiveproposedclm.feedback-interventions-backfire.rituals-aim-self

By Feedback Intervention Theory's classification, most standard leadership feedback rituals — numeric performance ratings, 360-degree appraisals, feedback sandwiches, praise-ratio targets — carry high self-cue and low task-cue loadings; this is a theory-based audit of practice argued by the theory's authors and practitioner syntheses, not a measured survey of organizations.

supportsreport-derivedSeed report, Key sources table
“360-degree appraisals "typically have design characteristics that reduce effectiveness"; feedback should stay task-focused and developmental use should be separated from administrative use”
DeNisi, A., Kluger, A. (2000). Feedback Effectiveness: Can 360-Degree Appraisals Be Improved?. Academy of Management Executive, 14(1), 129-139. doi:10.5465/AME.2000.2909845
contextualizesprimary-checkedAbstract (via PubMed 11125659)
“Idiosyncratic rater effects (62% and 53%) accounted for over half of the rating variance in both data sets.”
Scullen, S., Mount, M., Goff, M. (2000). Understanding the Latent Structure of Job Performance Ratings. Journal of Applied Psychology, 85(6), 956-970. doi:10.1037/0021-9010.85.6.956
contextualizesreport-derivedSeed report, Key sources table
“FSB–job performance ρ = 0.07 (k = 11, N = 1,910), credibility interval includes zero; "the relationship between FSB and performance was small"”
meta-analytic correlation of feedback-seeking behavior with job performance: ρ = 0.07, credibility interval includes zero (per seed report) (n = k = 11, N = 1,910)
Anseel, F., Beatty, A., Shen, W., Lievens, F., Sackett, P. (2015). How Are We Doing After 30 Years? A Meta-Analytic Review of the Antecedents and Outcomes of Feedback-Seeking Behavior. Journal of Management, 41(1), 318-348. doi:10.1177/0149206313484521
contextualizesreport-derivedSeed report, Key sources table
“"Feedback does not help employees thrive"; rater idiosyncrasy means ratings describe the rater; prescribes strengths-focused attention”
Buckingham, M., Goodall, A. (2019). The Feedback Fallacy. Harvard Business Review, March-April 2019. linknot peer-reviewed

Counter-evidence not yet searched.

strongproposedclm.feedback-interventions-backfire.goals-need-feedback

Feedback aimed at task standards reliably improves performance: goals plus feedback outperform goals alone (incremental d = 0.63 across 16 effect sizes), and the Cochrane review of audit-and-feedback (140 randomized trials) found a median +4.3 percentage-point absolute compliance improvement with an entirely positive interquartile range (0.5% to 16%), strongest with low baseline performance, credible sources, repetition, and explicit targets plus an action plan.

supportsprimary-checkedp. 708, Feedback section
“For goals to be effective, people need summary feedback that reveals progress in relation to their goals. If they do not know how they are doing, it is difficult or impossible for them to adjust the level or direction of their effort or to adjust their performance strategies to match what the goal requires.”
Locke, E., Latham, G. (2002). Building a Practically Useful Theory of Goal Setting and Task Motivation: A 35-Year Odyssey. American Psychologist, 57(9), 705-717. doi:10.1037/0003-066X.57.9.705
supportsprimary-checkedp. 708, Feedback section
“Summary feedback is a moderator of goal effects in that the combination of goals plus feedback is more effective than goals alone”
Locke, E., Latham, G. (2002). Building a Practically Useful Theory of Goal Setting and Task Motivation: A 35-Year Odyssey. American Psychologist, 57(9), 705-717. doi:10.1037/0003-066X.57.9.705
supportsprimary-checkedAbstract, Main results (via PubMed 22696318)
“After excluding studies at high risk of bias, there were 82 comparisons from 49 studies featuring dichotomous outcomes, and the weighted median adjusted RD was a 4.3% (interquartile range (IQR) 0.5% to 16%) absolute increase in healthcare professionals' compliance with desired practice.”
weighted median adjusted absolute risk difference in professional compliance: +4.3% (n = 82 low-risk-of-bias comparisons from 49 studies; 140 randomized trials included overall)
Ivers, N., Jamtvedt, G., Flottorp, S., Young, J., Odgaard-Jensen, J., French, S., O'Brien, M., Johansen, M., Grimshaw, J., Oxman, A. (2012). Audit and Feedback: Effects on Professional Practice and Healthcare Outcomes. Cochrane Database of Systematic Reviews, 2012(6), CD000259. doi:10.1002/14651858.CD000259.pub3
supportsprimary-checkedAbstract, Authors' conclusions (via PubMed 22696318)
“feedback may be more effective when baseline performance is low, the source is a supervisor or colleague, it is provided more than once, it is delivered in both verbal and written formats, and when it includes both explicit targets and an action plan”
Ivers, N., Jamtvedt, G., Flottorp, S., Young, J., Odgaard-Jensen, J., French, S., O'Brien, M., Johansen, M., Grimshaw, J., Oxman, A. (2012). Audit and Feedback: Effects on Professional Practice and Healthcare Outcomes. Cochrane Database of Systematic Reviews, 2012(6), CD000259. doi:10.1002/14651858.CD000259.pub3
supportsreport-derivedSeed report, §Evidence landscape
“the incremental value of adding feedback to goals is d = 0.63 across 16 effect sizes”
incremental d of feedback added to goal setting: 0.63; roughly doubles on complex tasks (per seed report) (n = 16 effect sizes)
Neubert, M. (1998). The Value of Feedback and Goal Setting over Goal Setting Alone and Potential Moderators of This Effect: A Meta-Analysis. Human Performance, 11(4), 321-335. doi:10.1207/s15327043hup1104_2

Counter-evidence searched: Searched 2026-07-07: the 2025 Cochrane update (CD000259.pub4, 292 studies) remains positive but stresses high variability, notes that 56% of interventions bundled co-interventions (education, reminders), and that effects at the median are small — consistent with, not contrary to, this claim's scoping. The Kluger & DeNisi negative tail bounds the claim from the other side: these reliable gains are demonstrations of the task-focused branch of feedback, not of feedback in general.

moderateproposedclm.feedback-interventions-backfire.wise-feedback-conditions

Candid negative feedback improves performance when it is task-focused, tied to explicitly high standards, and delivered inside a trusted relationship: a one-sentence 'wise feedback' note raised essay-revision rates among Black seventh-graders from 17% to 71% (covariate-adjusted; raw 27% vs 64%, p = .045, cells of ~11), experts actively seek the negative feedback novices avoid, and managers who discussed their upward feedback improved most.

supportsprimary-checkedStudy 1, p. 809
“The wise feedback treatment note stated, "I'm giving you these comments because I have very high expectations and I know that you can reach them." By contrast, the placebo control note stated, "I'm giving you these comments so that you'll have feedback on your paper."”
Yeager, D., Purdie-Vaughns, V., Garcia, J., Apfel, N., Brzustoski, P., Master, A., Hessert, W., Williams, M., Cohen, G. (2014). Breaking the Cycle of Mistrust: Wise Interventions to Provide Critical Feedback Across the Racial Divide. Journal of Experimental Psychology: General, 143(2), 804-824. doi:10.1037/a0033906
supportsprimary-checkedStudy 1 Results, p. 810
“An estimated 71% of African American students who received the wise feedback note revised their essays, compared with 17% of students who received the control note, b = 2.57, χ2(1) = 3.91, p = .045, OR = 11.95 (values are covariate adjusted; raw percentages were 64% vs. 27%, respectively).”
essay revision rate, wise vs control note (African American students): 71% vs 17% covariate-adjusted (raw 64% vs 27%); OR = 11.95, p = .045 (n = ~11 students per cell (44 students in Study 1))
Yeager, D., Purdie-Vaughns, V., Garcia, J., Apfel, N., Brzustoski, P., Master, A., Hessert, W., Williams, M., Cohen, G. (2014). Breaking the Cycle of Mistrust: Wise Interventions to Provide Critical Feedback Across the Racial Divide. Journal of Experimental Psychology: General, 143(2), 804-824. doi:10.1037/a0033906
supportsreport-derivedSeed report, Key sources table
“Novices seek and respond to positive feedback; experts seek and respond to negative feedback”
Finkelstein, S., Fishbach, A. (2012). Tell Me What I Did Wrong: Experts Seek and Respond to Negative Feedback. Journal of Consumer Research, 39(1), 22-38. link
supportsreport-derivedSeed report, Key sources table
“Managers who met with direct reports to discuss feedback improved more; overall improvement moderate (d ≈ 0.51 per Heslin & Latham 2004)”
Walker, A., Smither, J. (1999). A Five-Year Study of Upward Feedback: What Managers Do with Their Results Matters. Personnel Psychology, 52(2), 393-423. doi:10.1111/j.1744-6570.1999.tb00166.x

Counter-evidence not yet searched.