We all grew up with Hans Christian Andersen stories such as The Ugly Duckling. A mother duck’s eggs hatch, and one young bird looks different from the others. The farmyard calls the bird ugly, treats it as an outsider, and makes it seem as though something is wrong with it. It leaves. Later, when it sees a group of swans, it recognizes its reflection among them and discovers that it is a swan, not an ugly duckling after all.
The story gives us a hopeful way to read an anomaly: what appears to be a defect may actually reveal that the category is wrong. In business, an unusual result might expose a mistaken assumption, an overlooked customer need, or a new use for a product. The exception could point toward an opportunity.
There is also a less comforting possibility. The unusual result may be a warning that a product is failing, a process has changed, or customers are having a problem that the average score conceals. A spike in complaints is not a hidden swan. It may be smoke from a fire.
Marketers should take both possibilities seriously. An anomaly is not automatically an opportunity or a threat. It is a signal that something has departed from expectation. Its meaning depends on what produced it and what the evidence shows. Reeves and Fuller (2022) describe an anomaly as something that violates a familiar mental model. Ruef and Birkhead (2024) distinguish outliers, which are extreme data points, from anomalies, which are exceptional patterns. A single unusual observation may be noise, but a pattern gives us a reason to ask why the current explanation did not anticipate it.
The positive reading
In the positive reading, the anomaly suggests that the company’s model of the customer, product, or market is incomplete. The signal reveals an unmet need, a new use, or a source of value hiding in an unexpected place.
A company selling diagnostic machines to hospitals noticed stronger sales in Manhattan. Its first explanation was that the local salespeople were better. But the team kept looking and found another difference: Manhattan was the only region with technicians permanently on site.
Managers had assigned the technicians there to avoid wasting paid hours in traffic. Their regular presence also brought them into close contact with hospital staff. They learned more about customers’ needs and became effective at supporting sales. The company expanded the approach and reported eight additional points of market share and a 25% boost in margins (Reeves & Fuller, 2022).
The sales result challenged the company’s first explanation. The technicians’ presence helped explain what the sales figures alone could not: service interactions were also producing commercial value. The first anomaly sent the team looking; the second showed them where to look. By solving an operational problem, the technicians were also helping the company understand and serve its customers.
Holiday Stationstores tested salads and healthier grab-and-go snacks in 20 Twin Cities stores, expecting them to appeal to young adults and women. The surprising result was that fresh, healthy options also appealed to many of Holiday’s core adult male customers. Sales climbed as the weather warmed, and the company rolled out the offer to about 150 more local stores, with a larger advertising campaign planned. The pilot revealed an unexpected audience for the products and gave Holiday a reason to market them more broadly (NACS/Coca-Cola Retailing Research Council, 2014).
A European gas-station chain measured customer satisfaction across more than 150 outlets and found one station far ahead of the rest and another near the bottom. To test whether the difference came from location or customer service, executives switched the two stations’ managers. The ratings soon reversed with them: the former top performer fell to the bottom, while the lowest-rated station rose to the top. The test suggested that the customer experience followed the managers, giving the chain a specific cause to investigate. The account does not say whether the chain applied the finding at other locations (Owen, 2018).
The business opportunity came from seeing the shoe as customers were using it, not only as its makers had intended it to be used. The holes were part of the product’s design, but the idea of turning them into a way to personalize the shoe came from a customer’s behaviour. Crocs recognized the potential and connected that behaviour to its product line, distribution, and brand.
That is the positive lesson for marketers: customers sometimes reveal a product’s possibilities through what they do with it. A workaround, an improvised accessory, or a use that appears in a small customer group may expose an unmet need or a new form of value. The signal is worth following when it points to a need that repeats, a group that can be understood, or an idea that can be tested.
Obviously, the marketer’s task is not to declare every unusual use a breakthrough. They should ask what the behaviour tells them. Are customers solving a problem the product leaves unresolved? Are they using an existing feature in a way that creates a benefit the company has not communicated? Is a small group expressing a preference that could matter to a broader market? Those questions turn surprise into a disciplined search for opportunity.
The negative reading
A negative anomaly can warn marketers that something in the business is weakening. Sales may be falling in certain locations, customer satisfaction may be declining for a particular product, or complaints may be rising after a change. The pattern gives the team a place to investigate and a chance to respond before the problem spreads.
Barravecchia et al. (2025) studied 100 weeks of digital customer feedback about an anonymized smartwatch. Around week 60, discussion of battery life spiked, and the associated reviews were largely negative. The company investigated and identified a firmware update as the primary cause of decreased battery performance. It contacted affected customers, developed and tested a corrective update, and the authors report a subsequent positive shift in customer satisfaction. The timing of the complaints helped the company connect the performance decline to the update and track the effect of its response.
The same approach applies when results weaken in some locations. Marketers can compare those locations with stronger ones, examining customer needs, product availability, service, pricing, local competition, and execution. After diagnosing the cause, they can improve the solution across locations and adapt its execution to local conditions.
Averages can conceal important signals. Overall satisfaction may remain healthy while a particular feature, customer group, location, or recent release creates serious frustration. A complaint that appears minor in aggregate can become meaningful when it rises sharply, clusters around a change, or comes from customers whose experience differs from the rest. Marketers need to identify what changed, who is affected, and which locations or products need attention.
Cui et al. (2024) show that the method used to detect anomalies can shape which customer reviews attract attention. In their analysis of Dyson vacuum reviews on Xiaohongshu, isolation forest and density-based cluster analysis were sensitive to review length. An autoencoder more consistently surfaced reviews that experts judged relevant to product innovation. For marketers, this raises a practical question: could the method itself be making certain customer signals easier to see than others?
The same anomaly can carry both meanings
The positive and negative readings are useful distinctions, but real situations can contain both. A customer workaround may reveal a new use and expose a product shortcoming at the same time. A group’s unusual behaviour may show a promising segment while also revealing that the company’s current offer does not serve that group well.
The central question is what the signal reveals about the relationship between customers and the product. An unusual behaviour may point to creativity, friction, or both. If customers are adapting a product, the adaptation could point to creativity, friction, or both. If they are complaining about an unexpected feature, the comments could signal a defect, a mismatch between promise and experience, or a change in expectations.
A useful investigation begins by naming the expectation that was violated. “Customers in this segment should use the product in the standard way” is an assumption that can be examined. “This feedback is an outlier” may be a label that ends the inquiry too soon. An outlier is an unusual observation; an anomaly is a pattern whose meaning depends on why it departed from expectations (Ruef & Birkhead, 2024).
Next, look for what else changed around the signal. Check timing, product versions, customer groups, channels, locations, service interactions, and data quality. In Manhattan, the second clue was the technicians’ permanent presence. In the smartwatch case, the timing of the complaints helped investigators identify the firmware update. Context gave each signal a plausible explanation.
Then decide what evidence would separate competing explanations. If the anomaly suggests an unmet need, look for the same behaviour elsewhere, ask customers what they are trying to accomplish, or test a small version of the proposed solution. If it suggests a product failure, compare feedback from before and after a change, reproduce the problem, and check whether the issue is concentrated in a particular version or group. A company should not rewrite its strategy around a single surprising observation. It should use that observation to design a better investigation.
Organizations also have to decide where to direct attention. Cai and Canales (2024) study how combinations of organizational and market conditions shape firms’ attention to exploitation, exploration, or both. Their research is not about anomaly detection. It does, however, offer a useful lens on why signals can be missed: organizations allocate attention within existing priorities and conditions. A company focused on efficiency may be quick to classify an unusual customer as an exception. A company focused on exploration may be more willing to investigate, but it still needs a way to test what it finds. Recognizing an anomaly takes attention; learning from it takes time, resources, and permission to question familiar explanations.
Returning to the ugly duckling, what marketers should do when the farmyard notices a different bird. When something breaks expectations, ask these three questions:
What rule is the farmyard using to call this strange? Make the assumption visible. “Every region should sell in proportion to its sales team” can be investigated. “Those customers are unusual” can quietly dismiss them.
What else is different where the signal appears? Look at the surrounding conditions, including timing, people, product versions, channels, and customer needs. The apparent anomaly may make sense once the missing context comes into view.
What evidence would distinguish an opportunity from a warning? Look for a repeating pattern, a cause that fits the circumstances, or a small test that separates competing explanations. A useful signal should lead to something the company can verify.
An anomaly can reveal a swan, an overlooked customer opportunity, a faulty data point, a temporary shift, or a product defect that needs urgent correction. Investigation helps marketers determine which explanation fits. Andersen’s story reminds us that categories shape how people interpret difference. In business, a judgment made too early can cause a team to dismiss a promising signal as irrelevant or overlook an emerging warning as ordinary variation.
An anomaly can be an invitation to see a market differently or an early warning that the current offer is failing. The skill is to resist deciding which one before the evidence arrives.
What has your organization trained itself to ignore because noticing it would make the current model harder to defend?
References
Barravecchia, F., Mastrogiacomo, L., & Franceschini, F. (2025). Detecting digital voice of customer anomalies to improve product quality tracking. International Journal of Quality & Reliability Management. Advance online publication. https://doi.org/10.1108/IJQRM-07-2024-0229
Cai, J., & Canales, J. I. (2024). Attention focus and attention framework: A configuration perspective of attention to innovation. British Journal of Management, 35(2), 914–931. https://doi.org/10.1111/1467-8551.12744
Cui, X., Zhu, Z., Liu, L., Zhou, Q., & Liu, Q. (2024). Anomaly detection in consumer review analytics for idea generation in product innovation: Comparing machine learning and deep learning techniques. Technovation, 134, Article 103028. https://doi.org/10.1016/j.technovation.2024.103028
NACS/Coca-Cola Retailing Research Council. (2014, March). Get in the game: A report on how c-store operators are using the Playbook for Success to grow their businesses. https://www.ccrrc.org/report-directory/get-in-the-game
Owen, D. (2018, January 29). Customer satisfaction at the push of a button. The New Yorker. https://www.newyorker.com/magazine/2018/02/05/customer-satisfaction-at-the-push-of-a-button
Reeves, M., & Fuller, J. (2022, Winter). The imagination machine: How surprise triggers imagination. Rotman Management, 7–12.
Ruef, M., & Birkhead, C. (2024). Learning from outliers and anomalies. Academy of Management Perspectives. Advance online publication. https://doi.org/10.5465/amp.2023.0481
Leave a Reply