Why should we avoid noise?
The fact that people make different decisions in similar situations is not, in itself, unnatural or problematic. We are individuals with personal preferences and perspectives. If five movie critics watch the same film, we do not expect them to share the same opinion afterward. Variation in that type of setting is normal and even desirable.
The real problems begin when individual preferences overpower what should be shared, professional judgment criteria.
It does not create major consequences, except perhaps for the filmmakers, when one reviewer gives a movie 1 out of 6 stars while another gives it 6. But imagine that same kind of inconsistency happening in a hospital, in a courtroom, in child protection services, or among those deciding the insurance payout after your house has burned down. That is the type of noise we neither expect nor want.
Noise has a financial cost
One insurance company discovered that inconsistency in employee judgment was costing the business hundreds of millions of dollars. They already knew there would be some level of variation in decision-making, and they were comfortable with that. What came as a shock was how large the differences actually were.
After conducting a noise audit, they found that the average deviation in decisions made by claims handlers was as high as 55(!)%. In practice, this meant that one employee might award 9,500 dollars in a claim, while another, reviewing the exact same case, might award up to 16,000 dollars. This was not an extreme outlier, but the average level of variation. When comparing just two randomly selected employees, the gap could be even greater.
Similar findings were reported in departments responsible for setting customer premiums. In other words, noise was costly on both sides of the business.
Noise can cost lives
In the medical field, there has been significant focus on reducing noise. Still, large inconsistencies are often found in how decisions are made by healthcare professionals — even in identical medical cases.
Several American studies examined how pathologists assessed early signs of melanoma, a serious form of skin cancer. One study found that eight pathologists reviewing the same medical cases only agreed fully or partially in 62 percent of them. Another study showed that skin cancer was correctly diagnosed in just 64 percent of cases. This meant that doctors misdiagnosed approximately one-third of the cases.
The implications are serious. If this kind of variation is not detected and addressed, noise can lead to dangerous consequences. And once we know noise exists, the real question becomes what can we do to reduce it.

Different types of noise
The book outlines three distinct forms of noise: level noise, pattern noise, and occasion noise. These can appear separately or together, and each can significantly affect the outcome of a decision. Collectively, they are referred to as system noise.
Level noise refers to the individual differences in judgment based on who we are. This can include political beliefs, upbringing, life experience, or social influences. For instance, some judges may consistently hand down stricter sentences than their peers, while others are generally more lenient. In this case, error is defined as the deviation from the average decision.
Pattern noise occurs when level noise is inconsistent. This means a judge who is strict in one case may be unusually lenient in another. These inconsistencies may reflect a judge's personal philosophy about punishment. In other cases, they may be based on irrelevant emotional cues, such as the defendant resembling someone from the judge’s past.
Occasion noise is often the most unpredictable form of noise. It occurs randomly and can be difficult to detect, as it does not always appear the same way twice. This type of noise is caused by external factors in the decision-maker’s environment that affect outcomes. For example, a judge’s mood might be influenced by bad weather, a missed lunch, or their favorite football team losing the night before.
How to identify and reduce noise
It is important to understand that variation is natural whenever multiple people are involved in making decisions. A noise audit is a method used to identify the level of variation — in other words, to measure how much noise is present. The goal is to determine how much variation exists among employees who are performing the same tasks or solving similar problems, and then develop methods to reduce that variation.
Here is how to approach it:
Set up a project group and define roles clearly. Determine who will conduct the test, who will participate, who will represent the organization commissioning the test, and who will analyze and conclude based on the results. This often includes management, project leaders, and those carrying out the assessment.
Create a realistic case scenario to test on employees. This could involve an insurance claim, a cost estimate, an order of building materials, or similar decisions. The key is to make the cases as realistic and specific as possible so that measurements can be accurate and comparable to real-world decisions.
Involve leadership before the test and ask them to predict how much variation they expect in employee responses. Ask them to define acceptable levels of variance and estimate the financial consequences of incorrect assessments.
Let the project stakeholders write down anonymous hypotheses about the results they anticipate. This helps set realistic expectations and ensures accountability for the conclusions drawn from the test.
It is essential to remember that whenever people are making decisions, some variation will always exist. The key question is how much variation is actually present, and how much your organization can reasonably tolerate. Once you understand that, you can start making informed choices about how to reduce noise in a way that supports better outcomes and long-term profitability.
