# Feedback and Continuous Improvement | Siempre Feedback

Source: https://siemprefeedback.com/blog/feedback-continuous-improvement-research

*2026-10-14 by Miha*

On a Toyota assembly line, any worker who spots a problem can pull a cord (or press a button) above their station. A light comes on, a team leader walks over, and if the issue can't be fixed within the work cycle, the line stops. To an outsider that looks expensive. To Toyota it was the cheapest moment to learn something: the problem was small, fresh and visible to the person best placed to describe it.

That idea, that organisations get better by shortening the distance between something happening and someone learning from it, sits underneath almost every serious approach to quality, learning and service. This article looks at where it comes from, what the research actually says about feedback (including the uncomfortable parts), and the principles that separate feedback loops that compound from those that quietly waste everyone's time.

## What Feedback Actually Is

The word comes from engineering before it came to management. In its general sense, [feedback](https://en.wikipedia.org/wiki/Feedback) is what happens when the output of a system is routed back as an input, so that the system can adjust itself. A thermostat is the textbook case: it measures the room, compares the reading with a target, and switches the heating on or off to close the gap.

In 1948 the mathematician [Norbert Wiener](https://en.wikipedia.org/wiki/Norbert_Wiener) generalised the idea in *Cybernetics*, arguing that machines, organisms and organisations all regulate themselves through the same basic loop of action, measurement and correction. The field of [cybernetics](https://en.wikipedia.org/wiki/Cybernetics) that grew from his work drew a useful distinction. [Negative feedback](https://en.wikipedia.org/wiki/Negative_feedback) dampens deviation and keeps a system stable (the thermostat). Positive feedback amplifies it (a microphone squealing next to a speaker, or a bad review that leads to fewer customers, less revenue and worse service).

For any team or business, the practical lesson is simple. A feedback loop has four parts: something happens, a signal is captured, the signal is compared with what you wanted, and someone acts. Remove any one of them and the loop is broken. Most organisations have plenty of signals; far fewer have all four parts connected.

## A Short History of Continuous Improvement

Modern quality management starts with [Walter A. Shewhart](https://en.wikipedia.org/wiki/Walter_A._Shewhart) at Bell Telephone Laboratories in the 1920s. Shewhart realised that every process varies, and that the job of a manager is to tell ordinary, built-in variation apart from signals that something specific has gone wrong. His [control chart](https://en.wikipedia.org/wiki/Control_chart) and the distinction between [common and special causes](https://en.wikipedia.org/wiki/Common_cause_and_special_cause_(statistics)) of variation are still taught today.

His student [W. Edwards Deming](https://en.wikipedia.org/wiki/W._Edwards_Deming) carried the ideas to Japan after the Second World War and popularised the improvement cycle now known as [PDCA](https://en.wikipedia.org/wiki/PDCA): Plan, Do, Check (or Study), Act. Each pass through the cycle is a small experiment. You change something, observe the result, keep what works and plan the next change. In *Out of the Crisis* (1986) Deming argued that improvement has to be constant and forever, not a project with an end date.

Japanese manufacturers turned that philosophy into daily practice. [Kaizen](https://en.wikipedia.org/wiki/Kaizen), roughly "change for the better", describes continuous improvement through many small steps owned by the people doing the work. Under Taiichi Ohno, the [Toyota Production System](https://en.wikipedia.org/wiki/Toyota_Production_System) built feedback directly into the line through [jidoka](https://en.wikipedia.org/wiki/Autonomation) (machines and people stop when something is abnormal) and the [andon](https://en.wikipedia.org/wiki/Andon_(manufacturing)) signal described in the introduction. Those ideas later spread worldwide as [lean manufacturing](https://en.wikipedia.org/wiki/Lean_manufacturing) and influenced [Six Sigma](https://en.wikipedia.org/wiki/Six_Sigma).

The same logic shows up well beyond factories. Military strategist John Boyd's [OODA loop](https://en.wikipedia.org/wiki/OODA_loop) (observe, orient, decide, act) argues that whoever cycles through it faster gains the advantage. Agile software teams ship in short iterations for the same reason. And continual improvement is now a formal requirement in the [ISO 9000](https://en.wikipedia.org/wiki/ISO_9000) family of quality standards. The general name for all of this is the [continual improvement process](https://en.wikipedia.org/wiki/Continual_improvement_process).

## What the Research Says About Feedback (Including the Bad News)

It is tempting to assume that more feedback always means better performance. The evidence is more interesting than that. In 1996 Avraham Kluger and Angelo DeNisi published a landmark [meta-analysis](https://en.wikipedia.org/wiki/Meta-analysis) in *Psychological Bulletin* covering 607 effect sizes and more than 23,000 observations. On average, feedback interventions did improve performance, with a moderate effect size (d = 0.41). But **in more than a third of cases, feedback made performance worse.**

Their explanation was about attention. Feedback that keeps people focused on the task (what happened, what to change) tends to help. Feedback that pulls attention towards the self (how good or bad a person is, how they compare with colleagues) tends to hurt, because energy goes into protecting self-image rather than improving the work.

Education research points the same way. In a widely cited 2007 review in the *Review of Educational Research*, John Hattie and Helen Timperley synthesised a dozen meta-analyses and found an average effect of 0.79, placing feedback among the most powerful influences on learning. The variation was huge, though: praise and rewards were among the weakest forms, while information about the task and how to do it better was among the strongest. A 2020 re-analysis by Wisniewski, Zierer and Hattie, using a larger set of primary studies, found a more modest but still meaningful average effect (d = 0.48), with the same pattern: the more information the feedback contains, the more it helps.

Hattie and Timperley summarised useful feedback as answering three questions. Where am I going? How am I going? Where to next? Those questions work just as well for a hotel, a hospital ward or a delivery route as for a classroom.

## Why Timing Matters: Memory Rewrites Experience

Feedback collected long after an event is not a recording of the event. It is a reconstruction, and human memory edits as it reconstructs. Daniel Kahneman and colleagues demonstrated this in a 1993 experiment in which participants held a hand in painfully cold water. Many preferred to repeat a longer trial that ended slightly less cold over a shorter one, even though the longer trial contained strictly more discomfort. Retrospective judgements are dominated by the most intense moment and the final moment, a pattern now known as the [peak–end rule](https://en.wikipedia.org/wiki/Peak%E2%80%93end_rule).

Other well-documented effects pull in the same direction. [Recall bias](https://en.wikipedia.org/wiki/Recall_bias) distorts what people remember as time passes. The [negativity bias](https://en.wikipedia.org/wiki/Negativity_bias), reviewed by Roy Baumeister and colleagues in a 2001 paper with the memorable title "Bad Is Stronger Than Good", shows that negative events and negative feedback tend to carry more psychological weight than positive ones of equal size.

Researchers who need accurate data about everyday experience take these problems seriously. The [experience sampling method](https://en.wikipedia.org/wiki/Experience_sampling_method), also called ecological momentary assessment, asks people to report how they feel or what they are doing in the moment, precisely because in-the-moment reports are less distorted than recollections. The lesson for anyone running a service is that the closer a signal is captured to the experience it describes, the more it tells you about the experience and the less it tells you about the memory.

## Who Answers? The Silent-Majority Problem

A feedback loop is only as good as the sample flowing through it. [Participation bias](https://en.wikipedia.org/wiki/Participation_bias), also called non-response bias, occurs when the people who respond differ systematically from those who don't. The classic illustration is the 1936 US presidential poll by [The Literary Digest](https://en.wikipedia.org/wiki/The_Literary_Digest): around 2.4 million returned questionnaires predicted a clear win for Alf Landon, who then lost in a landslide to Franklin D. Roosevelt. Later analyses pointed to non-response as the main source of the error.

Interestingly, a high response rate alone does not guarantee an unbiased picture. A 2008 meta-analysis by Robert Groves and Emilia Peytcheva in *Public Opinion Quarterly* found that response rates were only a weak predictor of non-response bias. What matters more is *who* is missing. When feedback depends on effort (a long form, an account, a follow-up email days later), it tends to over-represent the very delighted, the very annoyed and the very patient, a form of [self-selection bias](https://en.wikipedia.org/wiki/Self-selection_bias).

The practical implication is to design feedback so that the quiet middle can respond as easily as the vocal extremes: short, optional, anonymous where possible, and available at the moment people are already thinking about the experience.

## Feedback, Recovery and Loyalty

In 1970 the economist Albert O. Hirschman described two ways people respond when an organisation disappoints them in [*Exit, Voice, and Loyalty*](https://en.wikipedia.org/wiki/Exit,_Voice,_and_Loyalty): they can leave (exit), or they can speak up (voice). Voice is the more useful of the two for the organisation, because it carries information and leaves room to fix things. When voicing a complaint is hard, more people simply exit.

Exit is expensive. In a 1990 *Harvard Business Review* article, Frederick Reichheld and W. Earl Sasser reported that in the service businesses they studied, reducing customer defections by 5% raised profits by between 25% and 85%. (The figure is often misquoted as "25 to 95%", and some researchers have since cautioned that the gains depend heavily on which customers are retained, but the direction of the finding is widely accepted.) This is why [customer retention](https://en.wikipedia.org/wiki/Customer_retention) and the wider discipline of [voice of the customer](https://en.wikipedia.org/wiki/Voice_of_the_customer) became core management topics.

Voice also creates a chance to recover. The [service recovery paradox](https://en.wikipedia.org/wiki/Service_recovery_paradox) holds that a customer whose problem is resolved well can end up more satisfied than one who never had a problem. The evidence deserves a careful reading. A 2007 meta-analysis by de Matos, Henrique and Rossi in the *Journal of Service Research* found the effect was real for satisfaction, but not significant for repurchase intentions, word of mouth or corporate image. Other studies suggest it appears mainly when failures are minor and the recovery is fast and excellent. The sensible conclusion is not to engineer failures, but that a problem caught and fixed quickly is far cheaper than one discovered later, after the customer has already decided.

## Culture: Feedback Only Flows Where It Is Safe

Many feedback problems are not technical at all. In a mid-1990s study of hospital units, Harvard researcher Amy Edmondson expected the best-led teams to report fewer medication errors. They reported more. On closer inspection, those teams were not making more mistakes; their members felt safe enough to report them. In a follow-up 1999 study of 51 work teams published in *Administrative Science Quarterly*, she introduced the concept of team [psychological safety](https://en.wikipedia.org/wiki/Psychological_safety), a shared belief that it is safe to take interpersonal risks, and showed it was linked to learning behaviour, which in turn was linked to performance.

The andon cord only works for the same reason. Accounts of the Toyota system stress that the team leader responds to a pull by helping, and often by thanking the person who raised the problem. Collecting feedback from customers or staff while punishing the bearers of bad news produces silence, not improvement. Leaders signal what is safe far more through how they react than through what they say.[/feedback-loops-growth-improvement/](https://siemprefeedback.com/feedback-loops-growth-improvement/)

## When Feedback Goes Wrong

Feedback loops can also be gamed. [Goodhart's law](https://en.wikipedia.org/wiki/Goodhart%27s_law), usually paraphrased as "when a measure becomes a target, it ceases to be a good measure", and the closely related [Campbell's law](https://en.wikipedia.org/wiki/Campbell%27s_law) describe what happens when a single score becomes the goal. Staff who are rewarded only for a satisfaction number learn to influence the number, sometimes by asking only happy customers or by nudging the rating, rather than by improving the experience.

Three other failure modes are common. The first is collecting without acting: people quickly notice when their input disappears, and stop giving it. The second is reacting to every single data point; Shewhart's distinction between common and special causes exists precisely to stop managers from chasing noise. The third is feedback aimed at people rather than processes, which Kluger and DeNisi's work suggests can lower performance rather than raise it.

## Principles for a Feedback Loop That Compounds

**Capture the signal close to the moment.** The peak–end rule and recall bias mean that delayed feedback measures memory more than experience.

1. **Make responding effortless.** Every extra field, login or day of delay filters out the quiet majority and skews the sample.
2. **Focus on the task, not the person.** Ask what happened and what would make it better, rather than asking people to grade individuals.
3. **Ask fewer, better questions.** One specific question at the right place beats twenty generic ones at the wrong time.
4. **Look for patterns before reacting.** Separate ordinary variation from genuine signals, in the spirit of Shewhart's control charts.
5. **Close the loop visibly.** Tell people what changed because of their input; it is the strongest encouragement to keep giving it.
6. **Treat scores as instruments, not targets.** Watch for Goodhart's law whenever a number is tied to rewards.
7. **Protect the messenger.** Psychological safety is the precondition for honest feedback from staff and customers alike.
8. **Repeat the cycle.** PDCA is a loop, not a line. Small improvements made every week outpace large ones made once a year.

## Conclusion: Small Loops, Repeated

The common thread from Wiener's thermostats to Deming's cycle, Toyota's andon cord and Edmondson's hospital wards is that improvement is less about brilliant insight than about well-designed loops: a fresh, honest signal, a team that feels safe to surface it, and a habit of acting on it quickly and then checking again.

The research also offers a healthy dose of humility. Feedback is powerful, but it is not automatically good. It works best when it is timely, specific, focused on the work, drawn from everyone rather than the loudest few, and followed by visible action. Get those conditions right and each small loop builds on the last, which is exactly what kaizen promised in the first place.

**Further Reading**

- [Why Feedback Loops Are Essential for Growth](https://siemprefeedback.com/feedback-loops-growth-improvement/)
- [What Is Feedback and Why Is It Important?](https://siemprefeedback.com/what-is-feedback/)
- [How to Capture In-the-Moment Feedback from Customers](https://siemprefeedback.com/blog/how-to-capture-in-the-moment-feedback-from-customers/)

**References**

Baumeister, R. F., Bratslavsky, E., Finkenauer, C., & Vohs, K. D. (2001). Bad is stronger than good. Review of General Psychology, 5(4), 323–370. [https://doi.org/10.1037/1089-2680.5.4.323](https://doi.org/10.1037/1089-2680.5.4.323)

de Matos, C. A., Henrique, J. L., & Rossi, C. A. V. (2007). Service recovery paradox: A meta-analysis. Journal of Service Research, 10(1), 60–77. [https://doi.org/10.1177/1094670507303012](https://doi.org/10.1177/1094670507303012)

Deming, W. E. (1986). Out of the Crisis. Cambridge, MA: MIT Press.

Edmondson, A. C. (1996). Learning from mistakes is easier said than done: Group and organizational influences on the detection and correction of human error. Journal of Applied Behavioral Science, 32(1), 5–28. [https://doi.org/10.1177/0021886396321001](https://doi.org/10.1177/0021886396321001)

Edmondson, A. (1999). Psychological safety and learning behavior in work teams. Administrative Science Quarterly, 44(2), 350–383. [https://doi.org/10.2307/2666999](https://doi.org/10.2307/2666999)

Groves, R. M., & Peytcheva, E. (2008). The impact of nonresponse rates on nonresponse bias: A meta-analysis. Public Opinion Quarterly, 72(2), 167–189. [https://doi.org/10.1093/poq/nfn011](https://doi.org/10.1093/poq/nfn011)

Hart, C. W. L., Heskett, J. L., & Sasser, W. E. (1990). The profitable art of service recovery. Harvard Business Review, 68(4), 148–156. [https://hbr.org/1990/07/the-profitable-art-of-service-recovery](https://hbr.org/1990/07/the-profitable-art-of-service-recovery)

Hattie, J., & Timperley, H. (2007). The power of feedback. Review of Educational Research, 77(1), 81–112. [https://doi.org/10.3102/003465430298487](https://doi.org/10.3102/003465430298487)

Hirschman, A. O. (1970). Exit, Voice, and Loyalty: Responses to Decline in Firms, Organizations, and States. Cambridge, MA: Harvard University Press.

Imai, M. (1986). Kaizen: The Key to Japan's Competitive Success. New York: McGraw-Hill.

Kahneman, D., Fredrickson, B. L., Schreiber, C. A., & Redelmeier, D. A. (1993). When more pain is preferred to less: Adding a better end. Psychological Science, 4(6), 401–405. [https://doi.org/10.1111/j.1467-9280.1993.tb00589.x](https://doi.org/10.1111/j.1467-9280.1993.tb00589.x)

Kluger, A. N., & 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. [https://doi.org/10.1037/0033-2909.119.2.254](https://doi.org/10.1037/0033-2909.119.2.254)

Ohno, T. (1988). Toyota Production System: Beyond Large-Scale Production. Portland, OR: Productivity Press.

Reichheld, F. F., & Sasser, W. E. (1990). Zero defections: Quality comes to services. Harvard Business Review, 68(5), 105–111. [https://hbr.org/1990/09/zero-defections-quality-comes-to-services](https://hbr.org/1990/09/zero-defections-quality-comes-to-services)

Wiener, N. (1948). Cybernetics: Or Control and Communication in the Animal and the Machine. Cambridge, MA: MIT Press.

Wisniewski, B., Zierer, K., & Hattie, J. (2020). The power of feedback revisited: A meta-analysis of educational feedback research. Frontiers in Psychology, 10, 3087. [https://doi.org/10.3389/fpsyg.2019.03087](https://doi.org/10.3389/fpsyg.2019.03087)