The most influential management concept of 2025: Noise


Every now and again, I come across a book which fundamentally changes the way I look at management.

And when I started to understand noise, it has been nothing short of revolutionary, because it flips a common management “principle” on its head:

Your teams, including management, are constantly making ineffective judgement decisions due to arbitrary factors, and many decisions would be more effective if people did not make them at all

A bold claim, I know, but one which the evidence of noise made me believe.

In their thought-provoking 2022 book Noise: A Flaw in Human Judgment, Daniel Kahneman (the creator of Thinking Fast and Slow), Olivier Sibony, and Cass R. Sunstein bring to light a pervasive issue in decision-making: the inconsistency of human judgment, or what they call “noise.”

An analogy to help explain it is to think of a target, and your aim for a decision is to hit the “bullseye” for the best decision for the business or situation.

In some cases, you or the team members will end up making a decision which really is ideal, essentially hitting the center of the bullseye.

The Problem: Bias vs. Noise

Bias—systematic errors that push decisions in a specific direction—has long been recognized as a flaw in human reasoning. It can be easy to understand and analyse because it’s directional. For example, if a hiring manager favors candidates from a specific university, or a department manager only ever accepts project proposals which have revenue potential above $10 million, their decisions are predictably skewed. Bias is systematic and can be measured, making it a familiar target for improvement efforts.

Bias when trying to hit a target might look like all the decisions missing the bullseye, but clustered together because they are based on similar reasoning and starting inputs.

Noise is an equally detrimental but less understood problem. Noise refers to the random variability in decisions about the same data and situations, made by different people or by the same person under different circumstances. Noise creates unnecessary and often invisible disparities in areas like hiring, medical diagnoses, criminal sentencing, and insurance underwriting. In your companies, it may manifest in certain types of projects (especially innovation projects) being accepted on one day, but rejected on another.

Noise would look like a target where the hits are all over the place, as different people make wildly different decisions, even if they also all using the exact same information.

Noise is also harder to detect than bias. It emerges when decisions vary unnecessarily across individuals or situations. Imagine two doctors diagnosing the same patient differently or two judges assigning vastly different sentences for similar crimes. These inconsistencies aren’t caused by systematic factors but by randomness in how judgments are made. Kahneman and his co-authors categorize noise into three types:

  1. Level Noise: Variability in the average leniency or strictness of different decision-makers (e.g., some judges are naturally harsher than others).
  2. Pattern Noise: Differences in how individuals respond to specific cases (e.g., one doctor might focus on symptoms A and B while another emphasizes symptom C).
  3. Occasion Noise: Fluctuations in decisions made by the same person at different times, influenced by factors like mood, fatigue, or even the weather.

What makes noise so insidious is its invisibility. Noise is unknown to most organisations is because often, the decisions or judgements of different people are never compared to one another. While bias often manifests as glaring patterns, noise is scattered around the organisation and harder to trace.

Noise also continues to be unknown because of the various KPIs used in departments and companies often rely on the summarising of end-results, into KPIs like average revenue per salesperson, total department costs per year, total number of ideas generated, average churn rate. By adding and averaging the individual results, and only looking at overall summary numbers, the differences can disappear. In fact, departments often do not realise there is a problem, because if you take the average of all the decisions, the overall results may actually look positive for the company as it looks like targets are being met.

This makes it challenging to address, even for those aware of its existence. In Noise, the authors provide striking examples of noise in two fields: criminal justice and insurance underwriting.

  1. Judges and Sentencing: The book highlights studies where judges were given identical case files but delivered significantly different sentences, sometimes diverging by several years. For example, one judge might sentence a defendant to probation, while another, reviewing the same facts, might impose a lengthy prison term of several years. Such variability reflects both level noise (differences in overall severity between judges) and pattern noise (different interpretations of case details). Occasion noise also plays a role, as decisions can shift based on external factors like the time of day or the judge’s mood.
  2. Insurance Underwriting: In the insurance industry, noise manifests in the inconsistent pricing of policies. Two underwriters evaluating the same risk might propose premiums that differ by 50% or more. These discrepancies arise not from biases in favor of or against specific applicants but from random judgment errors. Such noise leads to inefficiencies and potential financial losses, as overpricing can drive customers away, and underpricing erodes profitability.

The first example shows where Noise can affect fairness between different situations. The second one clearly shows where Noise leads to inefficiency in processes.

Example: Disparities in Prison Sentencing

In Noise, the authors delve into studies and statistics that vividly illustrate the problem of noise in judicial decisions. Here are some key examples and findings related to variability among judges:

  1. Sentencing Variation: Research conducted in U.S. courts shows that different judges often deliver vastly different sentences for similar crimes. For instance, one study found that the length of prison sentences for comparable offenses varied by an average of 19% between judges in the same jurisdiction. This means that a defendant’s fate depends as much on the judge they are assigned to as on the facts of their case.
  2. Time-of-Day Effects: Another striking example of noise involves “occasion noise.” A study of parole board decisions in Israel revealed that judges were more lenient after a meal break. Right before lunch or late in the day, the likelihood of a favorable parole decision dropped significantly, sometimes to 10%, compared to 65% immediately after the break. This highlights how external factors, such as fatigue or hunger, can randomly influence judicial outcomes.
  3. Judge Personality and Background: Judges’ individual characteristics also contribute to noise. For instance, studies in the U.S. found that judges appointed by different political parties tend to sentence defendants differently. Republican-appointed judges, on average, impose harsher sentences than their Democrat-appointed counterparts. While this reflects systemic bias, it also creates variability (or noise) when defendants with similar cases encounter different types of judges.
  4. Random Sentencing Studies: A well-known experiment involved giving identical case descriptions to multiple judges. Despite having the same information, the sentences ranged from probation to several years in prison. This dramatic spread highlights the unreliability of judgment in situations where consistency is critical.

As Kahneman and his co-authors emphasize, the random nature of these discrepancies reveals how deeply noise is embedded in human decision-making.

Bias and Noise working together against companies

In some situations, bias and noise may both work together to make companies and teams less effective.

If different teams, departments or locations are all using their own KPIs, and not comparing them against each other to become aware of differences, you may end up with “pockets” and silos forming within the company where everyone thinks their performance is good, but in fact none of them is ideal.

The major challenge is that people do not like to believe their decisions are affected by noise. They want to believe that their colleagues and team members would make consistent decisions based on the same information. After all, surely everyone is competent, otherwise they would not have a job at the same company.

Unfortunately, this is not the case.

Even worse, the easiest and most effective solution which Kahnemann and his colleagues found for noise was also the one which leadership were most likely to reject: replace a human needing to make a decision with a formula or algorithm to make it.

Now, this does not mean that people should be replaced by AI.

An algorithm in this case could be as simple an individual writing down a checklist: “If I see this data, as a rule I personally would make this decision”. And then the individual just following that “rule” they set for themselves.

Kahnemann showed that if they just followed the checklist, and others did it too, decisions would be significantly more consistent, effective and closer to the ideal outcome each time. The simple checklist outperformed human judgement in most cases.

The challenge is that even when people known what their own rule is, they are unlikely to follow it, still being swayed by their emotions and the situation each time, convincing themselves that this time they need to make an exception to the rule.

People have also shown the ironic view of liking the idea of other people using mechanical ways to make decisions, like checklists, but refusing to use it themselves. Their reasoning is that “Well, I would need to take into account all the various factors”.

Finally, humans believe that even if a mechanical or automatic way of making a decision works in 99.99% of cases to bring the “ideal” result, the fact that 0.01% of the time a person makes a better decision means that the mechanical system cannot be fully trusted, that it is “wrong”. Even if these results are orders of magnitude better than what a human would achieve. An example of this is with self-driving cars. Studies show that if all cars on the road were self-driving, fatalities and accidents would fall significantly. Well below the levels of accidents currently caused by human drivers. But the fact that self-driving cars would not result in ZERO accidents makes some people believe they could CAUSE and accident if they were to ride in them, making them think they should continue driving themselves. Even if the person driving themselves is in fact more likely to cause an accident than the car.

Why People Struggle with Noise and Bias

There are a number of reasons why people struggle to handle noise and bias.

  1. Illusion of Agreement: Many organizations assume that their professionals—whether judges, doctors, or executives—make consistent decisions. The illusion of agreement leads to overconfidence in the fairness and accuracy of judgment-based systems. The authors call this a “noise audit blind spot,” as most organizations fail to measure or even consider the presence of noise.
  2. Focus on Bias: Efforts to improve decision-making have historically focused on bias. For example, diversity training targets implicit biases, and standardized guidelines aim to prevent discrimination. While these are important, they’re only part of the solution. The emphasis on bias often overshadows the equally critical need to reduce noise.
  3. Resistance to Quantification: Measuring noise requires rigorous analysis, which often meets resistance. Decision-makers might view noise audits as a threat to their expertise, fearing that standardized procedures will reduce their autonomy or creativity. This cultural resistance undermines efforts to improve consistency.
  4. Underestimation of Randomness: People tend to underestimate how much randomness influences their decisions. Kahneman and his co-authors argue that decision-makers often believe their judgments are rational and objective, overlooking the subtle ways context, emotions, or irrelevant factors sway their choices.

Solutions: How to Tackle Noise and Bias

Kahneman and his co-authors propose several strategies to reduce noise and bias, emphasizing the need for systemic changes:

  1. Conduct Noise Audits: Organizations should measure the variability in their decision-making processes. By identifying where and how noise occurs, they can target specific areas for improvement.
  2. Introduce Decision Hygiene: Borrowing from the concept of hygiene in medicine, decision hygiene involves practices that minimize variability. These include using structured decision frameworks, breaking decisions into smaller, independent components, writing these as checklists and aggregating multiple judgments to reduce individual inconsistencies.
  3. Embrace Algorithms: Algorithms and statistical models are far more consistent than humans in many decision-making contexts. While not perfect, they can significantly reduce noise and bias when combined with human oversight.
  4. Debiasing Practices: While noise and bias are distinct, some practices can help with both. For instance, clearly defined criteria for decisions, transparency, and regular reviews of outcomes can improve judgment quality.

Why Processes to Reduce Noise and Bias Often Fail

So, we agree that noise and bias are problems. And we know some solutions to fix it. So why is it so hard for people to actually change

  1. Overreliance on Training: Training programs aimed at improving judgment rarely deliver lasting results. While they can raise awareness of issues like cognitive biases, they do little to eliminate noise. Training doesn’t change the fact that humans are inherently inconsistent decision-makers.
  2. Resistance to Standardization: Standardizing decision-making processes is one of the most effective ways to reduce noise, but it’s often met with resistance. Professionals may view algorithms, checklists, or structured decision frameworks as “mechanical” or “inhuman,” even when these tools outperform human judgment. This cultural resistance limits the adoption of effective noise-reduction strategies.
  3. Focus on Individual Errors: Organizations often focus on correcting individual mistakes rather than addressing systemic issues. Noise, however, is a systemic problem that requires systemic solutions, such as implementing decision rules or statistical models.
  4. Failure to Test Interventions: Efforts to improve decision-making frequently skip rigorous testing. Organizations may implement guidelines or tools without evaluating whether they reduce noise. Without robust testing, interventions can fail to produce meaningful improvements.
  5. The status quo bias itself: The ultimate irony is that biases themselves can keep people from wanting to improve their biases and reducing noise, even if they know it is the better solution. This includes the status quo bias, loss aversion, the anti creativity bias, the planning fallacy and many more.

Conclusion

Noise sheds light on an overlooked but critical flaw in human judgment. By understanding the dual threats of bias and noise, individuals and organizations can take more informed steps to improve decision-making processes. The path forward requires a shift in mindset: moving beyond individual judgment to embrace systemic solutions like noise audits, decision hygiene, and algorithmic assistance.

While addressing noise and bias is challenging, the potential benefits are enormous—greater fairness, efficiency, and accuracy in decisions that affect countless lives. The key, as Kahneman and his co-authors emphasize, is to view judgment not as an art but as a discipline that can and should be improved.

Idea to Value Podcast: Listen and Subscribe now

Listen and Subscribe to the Idea to Value Podcast. The best expert insights on Creativity and Innovation. If you like them, please leave us a review as well.
The following two tabs change content below.

Creativity & Innovation expert: I help individuals and companies build their creativity and innovation capabilities, so you can develop the next breakthrough idea which customers love. Chief Editor of Ideatovalue.com and Founder / CEO of Improvides Innovation Consulting. Coach / Speaker / Author / TEDx Speaker / Voted as one of the most influential innovation bloggers.

Latest posts by Nick Skillicorn (see all)

Leave a Reply

Your email address will not be published. Required fields are marked *