Feedback and Performance — Research and Evidence

Feedback is one of the most widely used — and most widely assumed-to-be-helpful — management tools, but the research evidence is more cautious than the popular framing. This page synthesises the science of feedback and of decision-making: when feedback improves performance and when it makes it worse, the principles that distinguish the two, the standing of common techniques such as the feedback sandwich, what people actually want, the value of seeking feedback rather than only giving it, the conditions under which feedback changes behaviour, and the firm-level cost of failing to decide.

The single most important finding governs everything below: feedback is not reliably helpful. On average it raises performance, but in a large minority of cases it lowers it, and the difference hinges on where the feedback directs the recipient’s attention. The confidence behind each finding varies — from a robust meta-analysis to a single, dated study — so each section below carries its own confidence label rather than a single page-level claim.

Does feedback actually improve performance, or can it backfire?

On average feedback improves performance, but it is not reliably helpful. Kluger and DeNisi’s 1996 meta-analysis — 607 effect sizes from 23,663 observations — found feedback raised performance by an average of d ≈ 0.41 (small-to-moderate), but over a third of effects were negative, meaning feedback actually reduced performance. In their words, feedback “improved performance on average (d = .41) but… over 1/3 of the FIs decreased performance. This finding cannot be explained by sampling error, feedback sign, or existing theories.” The harm was not explained by whether the feedback was positive or negative.

The proposed explanation is Feedback Intervention Theory. Feedback directs attention across a hierarchy — task learning, task motivation, and meta-task (self-related) processes. When feedback keeps attention on the task and how to do it better, it tends to help; when it pushes attention “up” toward the self — ego, identity, self-worth — it tends to hurt, because cognitive resources go to defending the self rather than improving the task. The more feedback is “about you” rather than “about the work,” the more likely it is to backfire.

The caveats matter. This is a 1996 meta-analysis spanning lab and field studies and many task types; the moderators are real but, the authors admit, “still poorly understood.” The d ≈ 0.41 average masks enormous variability — which is the point: the average benefit conceals frequent harm. The practical implication is that “we give lots of feedback” is not self-evidently good; poorly aimed, ego-directed feedback is a documented way to make performance worse, and feedback helps when it is aimed at the task, not the person.

Source: Kluger & DeNisi, The Effects of Feedback Interventions on Performance: A Historical Review, a Meta-Analysis, and a Preliminary Feedback Intervention Theory, Psychological Bulletin 119(2), pp. 254–284 (1996). Confidence: verified.

What actually makes feedback effective?

The best-supported principle is task-focus: feedback that keeps attention on the task and how to improve it tends to help, while feedback that directs attention to the self tends to harm. Feedback should also be specific and actionable rather than vague, and it is accepted and acted on far more readily when it comes from a credible, trusted source.

The task-focus principle follows directly from Kluger and DeNisi (1996): feedback that keeps attention on the task and how to improve it helps, while feedback that directs attention to the self tends to harm. Hattie and Timperley (2007), reviewing the educational literature, reach a compatible conclusion — feedback is “among the most powerful influences on learning,” but its effect “can be either positive or negative” depending on type and delivery, and threat to self-esteem cuts effectiveness sharply. Their framework holds that effective feedback answers three questions: Where am I going? How am I going? Where to next?

Specific and actionable beats vague. Generic praise carries little task information. And source credibility and trust matter for whether feedback is accepted at all: Ilgen, Fisher and Taylor (1979) established that feedback changes behaviour only through a chain — how it is perceived, whether it is accepted, and whether the recipient is willing to act — and that feedback from credible, trusted sources is acted on while feedback from distrusted sources is discounted. Identical words land differently from a respected manager than from a distrusted one.

Two qualifications are important. The timing question (immediate versus delayed feedback) is genuinely unsettled and appears to depend on task complexity. And a transfer caveat applies to the education research: Hattie and Timperley is education-sector research, and its large classroom effect sizes should not be read as workplace effects; the workplace anchors here are Kluger and DeNisi and Ilgen and colleagues. In practice, the principles combine: name the specific behaviour and its impact, tie it to the task and a concrete next step, deliver it from earned trust, and keep it off the person’s identity. Progress-toward-a-goal feedback — knowing where you stand against an expectation — is treated separately in the Role Clarity and Goal-Setting research.

Sources: Kluger & DeNisi, Psychological Bulletin 119(2), pp. 254–284 (1996); Hattie & Timperley, The Power of Feedback, Review of Educational Research 77(1) (2007) (education-sector — transfer with care); Ilgen, Fisher & Taylor, Consequences of Individual Feedback on Behavior in Organizations, Journal of Applied Psychology 64(4) (1979). Confidence: industry-consensus.

Is the feedback sandwich actually effective?

The praise–criticism–praise sandwich is one of the most commonly taught feedback techniques, yet it rests on remarkably little rigorous head-to-head testing, and the few controlled experiments are small and mixed. Its best-documented effect is improving the giver’s comfort and the recipient’s perception, not their performance.

The clearest empirical statement is Parkes and colleagues (2013): in two studies of written peer feedback among medical students, the technique is “commonly recommended… despite scant evidence of its efficacy,” and students “think feedback sandwiches positively impact subsequent performance when there is no evidence that they do” — that is, sandwiches affected perceptions but not performance.

The experimental record is genuinely mixed, and honesty requires flagging it. Prochazka and colleagues (2020), the most-cited controlled experiment on the sandwich (91 students, math problems), reported the opposite of the popular “sandwiches don’t work” claim: the sandwich group performed better on the next task. But it is small, lab-based, used generic praise unrelated to performance, and tested one task type — enough to prevent a blanket dismissal, not enough to establish superiority.

The proposed failure mechanism comes from Kluger and DeNisi (1996): mixed signals can shift attention toward the self, and the corrective core can be diluted or discounted through primacy and recency effects. Over time, recipients may learn that praise is merely a setup for criticism, eroding trust in the praise itself. The defensible position is therefore that the sandwich is not evidence-based as a reliably superior method; its best-documented effect is improving the giver’s comfort and the recipient’s perception, not their performance. Popular alternatives such as Radical Candor (“care personally and challenge directly”) are trade-book frameworks, not evidence, and should be labelled as such.

Sources: Prochazka, Ovcari & Durinik, Sandwich Feedback: The Empirical Evidence of Its Effectiveness, Learning and Motivation 71:101649 (2020); Parkes, Abercrombie & McCarty, Feedback Sandwiches Affect Perceptions but Not Performance, Advances in Health Sciences Education 18 (2013); Kluger & DeNisi, Psychological Bulletin 119(2) (1996); Scott, Radical Candor (St. Martin’s Press, 2017) (trade book, practitioner not evidence). Confidence: directional.

Do people actually want corrective feedback, or do they just say they do?

A widely cited consulting survey found that 57% of employees said they preferred corrective feedback to praise, but this is proprietary, non-peer-reviewed, stated-preference data, and rigorous research shows that what people say they want diverges from how they actually react to and act on negative feedback.

Sourced precisely: in their 2014 Harvard Business Review article, Zenger and Folkman reported that “57% preferred corrective feedback; only 43% preferred praise,” that “72% said they thought their performance would improve if their managers would provide corrective feedback,” and that 92% agreed negative feedback, delivered appropriately, improves performance (sample: 2,500+ employees). A separate, later dataset (8,715 respondents) reframed the question as “negative” versus “positive” and reported 66% preferring negative feedback. These are two different studies at different times with different wording — not two reports of one survey.

This requires a clear flag: it is proprietary consulting survey data, never peer-reviewed, generated with the firm’s own instruments, and every headline number is a stated preference or self-reported belief, not measured behaviour or performance. The 72% figure is respondents predicting their own improvement, not an observed change.

The rigorous nuance is that stated desire diverges from actual response. Self-evaluation research documents competing motives — wanting to feel good (and so discounting negative feedback) versus wanting accurate information. The strongest peer-reviewed evidence that preference is conditional is Finkelstein and Fishbach (2012), who across five studies found “a shift from positive to negative feedback as people gain expertise… novices sought and responded to positive feedback, and experts sought and responded to negative feedback.” So “people want corrective feedback” is broadly true but moderated by expertise, ego, and whether the trait feels changeable. The practical implication is that employees genuinely value honest correction more than managers assume, but one cannot infer from a survey that any given person will receive a specific piece of hard feedback well in the moment — wanting feedback and reacting well to it are different things.

Sources: Zenger & Folkman, Your Employees Want the Negative Feedback You Hate to Give, Harvard Business Review (2014) (underlying data proprietary, non-peer-reviewed); Finkelstein & Fishbach, Tell Me What I Did Wrong: Experts Seek and Respond to Negative Feedback, Journal of Consumer Research 39(1) (2012). Confidence: single-source.

Does seeking feedback improve performance and adaptation?

Actively seeking feedback is consistently linked to better adjustment, learning, and socialisation, but its direct link to measured performance is small. People weigh real ego and image costs against the informational value when deciding whether to ask.

Ashford and Cummings (1983) reframed feedback as an individual resource people proactively create — through monitoring (watching for cues) and inquiry (directly asking) — not just passively receive. Ashford, Blatt and VandeWalle (2003) organised two decades of research around three motives: instrumental (achieve a goal), ego-based (protect one’s ego), and image-based (protect one’s reputation). The central insight is a cost–benefit calculus: information has value, but asking carries an ego cost (the risk of hearing something unflattering) and an image cost (the risk of looking unsure).

On outcomes, the honest picture is mixed. Anseel and colleagues’ (2015) meta-analysis confirmed the framework — learning-goal orientation, high self-esteem, transformational leadership, and a high-quality leader relationship were positively associated with feedback-seeking, while tenure and age were negatively associated — but explicitly challenged two dominant views, finding “the relationship between uncertainty and FSB was negative and the relationship between FSB and performance was small.” So feedback-seeking robustly aids socialisation, role clarity, and adjustment, but its direct line to higher measured performance is weaker than the popular framing implies.

The practical levers that follow are clear: leaders who openly seek feedback model the behaviour, and the climate — trust, a learning orientation, and not punishing questions — strongly shapes whether people ask. The ego and image costs are real and worth deliberately lowering.

Sources: Anseel, Beatty, Shen, Lievens & Sackett, 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), pp. 318–348 (2015); Ashford & Cummings, Feedback as an Individual Resource: Personal Strategies of Creating Information, Organizational Behavior and Human Performance 32(3) (1983); Ashford, Blatt & VandeWalle, Reflections on the Looking Glass: A Review of Research on Feedback-Seeking Behavior in Organizations, Journal of Management 29(6) (2003). Confidence: industry-consensus.

Does a feedback orientation or feedback culture improve outcomes?

An individual’s receptivity to feedback and an organisation’s supportive feedback culture are theorised — and increasingly supported — to determine whether feedback changes behaviour, though the foundational model is conceptual and the empirical gains from feedback systems are, on average, small.

London and Smither (2002) introduced two constructs to explain why the same feedback lands differently in different organisations: feedback orientation (an individual’s overall receptivity — seeing feedback as useful, feeling accountable to act, being socially aware, and feeling self-assured in handling it) and feedback culture (the degree to which an organisation supports non-threatening, behaviourally focused feedback, provides coaching to interpret it, and links improvement to development and rewards). Their core claim is that feedback only improves performance when orientation and culture are favourable — the receiving conditions matter as much as the message.

The honest caveat is that London and Smither (2002) is a conceptual model, not an empirical test. The strongest empirical anchor on whether feedback systems change behaviour over time is Smither, London and Reilly’s (2005) meta-analysis of 24 longitudinal 360-degree feedback studies, which found “improvement… over time is generally small” — corrected effect sizes around d ≈ .15 for direct-report and supervisor ratings, near zero for self-ratings. Simply delivering feedback — even structured 360s — produces modest average gains, concentrated among recipients with a positive feedback orientation, in supportive cultures, who set goals and act.

The practical reframing is that building a culture where feedback is routine, behaviour-focused, and tied to support and follow-through does more than buying a feedback tool or running an annual 360. The active mechanism is the receiving environment, not the instrument.

Sources: Smither, London & Reilly, Does Performance Improve Following Multisource Feedback? A Theoretical Model, Meta-Analysis, and Review of Empirical Findings, Personnel Psychology 58(1), pp. 33–66 (2005); London & Smither, Feedback Orientation, Feedback Culture, and the Longitudinal Performance Management Process, Human Resource Management Review 12(1) (2002). Confidence: directional.

What does chronic indecision and slow decision-making cost an organisation?

A seminal but dated and context-bound study links faster strategic decision-making to better firm performance, but the evidence is limited, drawn from one volatile industry, and “fast” explicitly means well-informed, not rushed.

The firm-level cost of failing to decide and act rests largely on one body of work. Eisenhardt’s 1989 study, Making Fast Strategic Decisions in High-Velocity Environments, is the seminal source. Based on an inductive study of eight microcomputer firms, it found a link between decision speed and firm performance: fast deciders performed well, slow deciders did poorly or failed. Critically — and most recaps miss this — Eisenhardt found fast decision-makers used more real-time information, considered more alternatives simultaneously, and used experienced counsel and active conflict resolution. “Fast” meant well-structured and decisive, not rushed or under-analysed.

The caveats are substantial, which is why the finding is treated as single-source and directional. The 1989 result is (1) from a small, inductive, theory-building sample of eight firms — propositions generated, not statistically tested across many firms; (2) specific to a high-velocity 1980s industry; and (3) decades old, with limited comprehensive replication since. Linking decision speed to outcomes is genuinely context-dependent; the evidence does not support a universal “faster decisions = better results” law, and certainly not for every decision type.

The transferable insight is qualitative: chronic deferral of consequential calls — a needed pivot, a hire, an exit — carries real opportunity cost, and the remedy is decisiveness backed by good information, not impulsiveness. This is the evidentiary basis for the “for want of a horse” idea, that a small, timely decision can avert a large, compounding loss. The individual-level drivers of slow decisions — status-quo bias and omission bias — belong to the practice material; they are covered in the Difficult Conversations, Underperformance, and Termination guide, and the team-level consequences of deferral are set out in Toxic Behaviour and Tolerated Underperformance.

Sources: Eisenhardt, Making Fast Strategic Decisions in High-Velocity Environments, Academy of Management Journal 32(3), pp. 543–576 (1989); Bourgeois & Eisenhardt, Politics of Strategic Decision Making in High-Velocity Environments, Academy of Management Journal 31 (1988). Confidence: single-source.

---

This page is general information summarising published research, not advice for a specific situation. The findings above describe averages and tendencies across studies, with effect sizes and confidence labels stated so the strength of each can be weighed; how they apply to a particular team, role, or decision depends on the circumstances.