The Power of Anomaly

Idea in Brief The Problem Trends are well established by the time most companies pick up on them, and opportunities have been seized by competitors. Why It Happens Companies often don’t recognize weak early signals that a trend is emerging because managers are internally focused, unwilling to challenge their assumptions, and uncomfortable with ambiguity. The Solution Managers can employ new analytics tools to surface anomalies and identify the ones that offer opportunities. They should name and build a convincing narrative around those opportunities and be prepared to quickly test emerging trends and shape them in their favor. Identifying the next big thing is often treated as an exercise in analyzing trends. But that’s misleading. By the time a trend is established, any opportunities it presents have most likely already been captured by competitors. And although a company may need to reflect trends in its business plans, that can be more a matter of catching up with rivals than of gaining a competitive edge. To take advantage of emerging trends, companies must identify them when they are embryonic—not purely speculative, but not yet named or widely known. At that stage the signs will be merely anomalies: weak signals that are in some way surprising but not entirely clear in scope or import. Most anomalies don’t become meaningful trends, of course, but some do—and the businesses that identify and interpret them early will steal a march on the competition. Instagram’s founders, for example, focused their initially multifunctional social networking app, originally called Burbn, on image sharing after they noticed an unusual amount of activity around that feature. In this article we present a process for spotting anomalies that have the potential to drive a business, but it’s more than mechanical: Anomaly-driven strategy requires being open to unexpected ideas that may overturn long-held assumptions. Only if you are willing to look at your business from the outside in, question your existing models, and embrace ambiguity will you be able to identify the diamonds hiding in the data. Let’s look at what that process involves. An Open Mind We usually can’t see what we aren’t looking for, so anomalies are often missed. If we do see them, we may ignore them, because people and organizations are predisposed to confirmation bias, focusing on what aligns with their mental models rather than what violates them. Indeed, the word “anomaly” is most often used to dismiss a data point as unrepresentative and irrelevant. And even if we don’t ignore anomalies, we may not try to interpret or explore them. So managers must be prepared to: Take an external perspective. As companies grow, they often become more introverted. Managers disproportionately focus on internal processes rather than changes in the outside world. A company can be thought of like a sphere—as it expands, its internal volume increases faster than its external surface area does, driving internal focus by default. But anomalies that may signal new trends will appear in the external world. As Intel’s former CEO Andy Grove famously said, “When spring comes, snow melts first at the periphery, because that is where it is most exposed.” Thus managers need to adopt an external perspective. That involves understanding the behaviors not only of average customers but also of atypical customers, lapsed customers, noncustomers, and the broader consumer landscape. It involves monitoring both established competitors and new challengers that may have different models or come from entirely different industries. And it involves continually scanning for changes in the social, political, and technological context that might shape the emergence of new needs. Kenneth Libbrecht Companies can get a better sense of the external world by encouraging managers and employees to spend more time engaging directly with potential customers. They can also access data from new sources, such as social media text, to discern shifts in consumer sentiment or behavior. They can tap into new data-collection capabilities, such as sensors or customer-facing digital platforms, often by collaborating in multicompany ecosystems. And they can examine change on multiple timescales—from very short ones, where new deviations can be observed, to very long ones, which reveal important but slower-moving shifts. Challenge assumptions and mental models. Gaining new information is not enough to spot potential trends. We base our mental models of the world on what we can sense, but those models are not reality—they are just a representation of it. And they don’t automatically change when reality changes. We can easily get stuck in them and thus discount or ignore new information. For example, when given an opportunity to buy the patents for the telephone shortly after it was invented, William Orton, the president of Western Union (then the dominant telegraph company), reportedly dismissed the device as an “electrical toy.” The view prevailing at the time was that the telephone would be used merely for short updates—a purpose for which the telegraph would remain more practical. Orton’s mental model didn’t shift to allow for new possibilities that the telephone would unlock, such as extended conversations at a distance, until a couple of years later, when the trend of telephone adoption had become obvious to all. Identifying anomalies involves more than just having access to better new signals; it requires an ability to imagine the potential implications of those signals. We must recognize that models are just that—and then be flexible about updating them over time. That means questioning assumptions and asking what if the reality were otherwise; adopting multiple, perhaps even contradictory, models in order to test various angles; and thinking in analogies to widen the range of models available to us. Embrace ambiguity. Anomalies usually don’t point to obvious actions. They fall into gaps that are not well explained by current models, and the data supporting them may not be clear-cut. Established businesses often have difficulty embracing this kind of ambiguity, as is demonstrated by the adage “What gets measured gets managed.” But ambiguity creates opportunity. If a signal were clear, everyone would see and act on it, making it competitively neutral. As the military strategist Carl von Clausewitz wrote, “[Attempts] to equip the conduct of war with principles, rules, or even systems [that] considered only factors that could be mathematically calculated are objectionable…because they aim at fixed values. In war everything is uncertain and variable.” Similarly, to imagine and respond to trends before they are clear, leaders must act on uncertain and shifting information. That requires an altered approach to evaluating and seizing opportunities. A common error by large businesses is to require that any new initiative be backed by unambiguous return-on-investment analysis showing a clear course of action leading to a quantified payoff. But given the uncertainty and “path dependency” (whereby what has occurred in the past persists because of resistance to change) of emerging trends, such an analysis is rarely possible. As the entrepreneur and author Steve Blank has said, such an exercise is “like a divide by zero problem.” Ambiguity creates opportunity. If a signal were clear, everyone would see and act on it, making it competitively neutral. Leaders should be willing to evaluate nascent opportunities on other dimensions: Is there a compelling story you can tell in which an anomaly would become a big opportunity? Would the opportunity make strategic sense given your company’s capabilities? How could new assumptions be tested? Companies cannot stay on top forever with their existing offerings alone, so sustaining success requires acting on fuzzy opportunities. Let’s now consider how managers can best leverage their open-mindedness. The Anomaly-Driven Process In a typical new-product-development process, market researchers comb through historical data, form multiple hypotheses, and check them through experimentation. That allows the researchers to identify a value proposition, which can then be developed and rolled out. The process may take many months. Anomaly-driven innovation is very different. To begin with, market researchers don’t necessarily start with any hypotheses, and the data is analyzed as soon as it is generated. The process is less about predicting the future from the past than about making sense of the immediate present and using it to imagine possible futures. As a result, the timescale compresses to weeks, not months. Typically the process involves four steps. 1. Analyze and visualize granular data. Because averages smooth out differences, anomalies are likely to emerge only from granular data sources—those that are harvested with great frequency, are de-averaged for specific geographies, or report individual transactions. But inspecting all data at such a granular level isn’t feasible in large organizations, so the data must be structured in ways that allow anomalies to be visualized and investigated. New analytical tools can identify the underlying structure of large data sets, allowing them to be interpreted without relying on averages and summary statistics. Data such as consumers’ posts on social platforms or competitors’ investment announcements can be analyzed with semantic clustering algorithms, making the underlying patterns of belief or behavior explicit and interpretable. To find the anomalies within these patterns, visualize the data and interrogate it to identify outliers and gaps. Which clusters display growing or changing characteristics over time? 2. Identify the anomalies that matter. Large data sets are likely to include lots of anomalies, not all of which will be meaningful. To decide which ones should be acted on, identify those that are most likely to indicate something about the future—and that tell you something you don’t already know. By definition, anomalies are generally small when they’re spotted, but their potential can be assessed on three dimensions. (1) Momentum: Is the anomaly persistent over time? Is it growing rapidly? Do early adopters show passion for it? (2) Robustness: Can it be seen from different perspectives or in multiple data sources? Is it consistent with other shifts in the environment? (3) Impact: Does the anomaly point to opportunities to plug significant gaps in the landscape of current offerings? What might be the implications of it? Kenneth Libbrecht, a professor of physics at Caltech, studies the molecular dynamics of crystal growth and the physics of snowflakes. He photographs the fascinating variety of snowflake shapes and patterns and grows his own snow crystals with designs not found in nature. Kenneth Libbrecht In 2000, two years after the founding of Google, the company was considering expanding beyond text links into new types of search when it noticed a significant anomaly: Searches for the green dress worn by Jennifer Lopez at the Grammy Awards broke the record for the most popular query ever. Google had already identified image search as a potential opportunity, but the enormous number of searches for the dress indicated just how much momentum and potential impact the not-yet-existent service had, persuading Google to make it a priority and launch it the next year. 3. Give them a narrative and a name. The next step is to try to make sense of a promising anomaly by creating the story it would tell if it became a trend. This exercise helps in imagining futures, which can be tested against a broader knowledge of the changing environment. If it’s compelling, such a narrative can help spread the idea through your organization, enabling others to see the opportunity or the threat. To make sense of the narrative around an anomaly, ask yourself: What could be going on here? What could it lead to? What additional facts would be consistent with this story or challenge it? What is the belief underpinning this future that violates our existing beliefs? Is there a “missing piece” that could confirm or accelerate this story? What pattern of signals would validate it? An especially powerful part of storytelling is naming, which allows an unlabeled thought in one person’s mind to evolve into an idea that can be shared with a group. Names also help others in an organization understand and engage with new concepts by making brand-new ones (such as how the Apple Watch denotes a small, sleek computer on your wrist) feel more familiar, communicating a function (such as how an “automatic teller machine” withdraws money from banks), or capturing attention (“Amazon,” the name of the world’s biggest river, signaled the company’s ambitions for growth and breadth). 4. Probe, shape, and commit. Having identified and named a potential new trend, the final step is to act on it. Even important anomalies are weak signals, so testing an idea rapidly and iteratively—for instance, by running small experiments—will help you both validate the idea and learn more, by revealing other anomalies that might prompt you to rethink your idea. In a changing environment we learn, counterintuitively, by acting and then thinking. Acting on opportunities should involve more than learning and reacting, however. Because anomalies presage emerging opportunities, they tend to be malleable. Thus early actors can shape how they will unfold by developing new ideas in collaboration with customers and partners. For example, the 2003 SARS epidemic in Asia caused e-commerce adoption to accelerate, spurring online retailers to focus on consumer channels in addition to B2B sales. Because many customers mistrusted online purchases, Alibaba launched Alipay, which reduced transaction risk and rated sellers, increasing trust and enabling the company to become China’s e-commerce leader. In other words, it shaped the context for e-commerce and thereby expanded its potential. As the Alibaba example illustrates, when you’ve identified and validated a promising opportunity, you should back your hunch. Large organizations are too often unwilling to make more than minor adjustments to existing budgets. Companies that are prepared to shift capital allocation aggressively generally achieve stronger growth—especially after crises, when opportunities are disproportionately likely to emerge. Let’s look at one company that applied the principles and processes we’ve outlined to identify new growth opportunities. How a Semiconductor Supplier Discovered Biobanking Facing slowing growth in the late 2000s, Brooks Automation, a U.S. semiconductor equipment manufacturer, sought opportunities in adjacent markets that would leverage its core technological capabilities, such as creating very cold vacuum environments and handling wafers within them. It mapped a granular web of who was citing its patents directly and also who was citing the companies citing its patents. Six core technologies, including cryopumps and materials handling, had 90 first-level citations and nearly 400 second-level citations among them. The exhibit “Anomaly Spotting Using Patent Citations” shows that the companies citing the patents fell into five distinct application domains. As Brooks managers dug into the details, they identified an interesting and hitherto unrecognized pattern: The map showed not only direct links between their company’s patents for cryocooling and MRI and cryosurgery applications but also many weaker ties between other Brooks technologies, such as automation systems, and other medical applications, such as medical robotics, suggesting that medicine might be a promising area to explore. To test the possibilities, Brooks zoomed in on medical applications and discovered that although it held no patents in cryogenic tissue cooling, the cluster of its other patents around that topic was growing. And research indicated that existing approaches to freezing human tissue were haphazard and unreliable, suggesting a significant gap in the market that Brooks could potentially fill. Having narrowed down the opportunity, Brooks moved to the third step in our process. Through ideation sessions with engineers in its cryopump and robotic wafer manipulation businesses, the company developed a compelling story around this opportunity: that Brooks could foster and grow “precision biobanking” to support the life sciences market with digitization, better-controlled environments, and automatic retrieval of samples. To seize the opportunity before other competitors saw and acted on it, Brooks made a series of small, strategic acquisitions that added industry experience and filled capability gaps while also developing new products to store biological samples more reliably and systematically. Having entered the new niche early, Brooks became the market leader, and its life sciences business grew at 35% annually over the next decade. That allowed it to outperform its peers, which continued to focus largely on semiconductor customers, and to achieve annualized total shareholder returns of 22% during the 2010s. . . . The late Clay Christensen, whose theory of disruptive change redefined how managers think about innovation, hung a wooden sign outside his office at Harvard Business School. (An enthusiastic wood carver in his spare time, he had made the sign himself.) It read: Anomalies Wanted. Christensen delighted in finding anomalies, because he recognized that meeting the unexpected leads to the greatest advances in knowledge. Anomalies inspire us to test and improve and perhaps change our theories and models of how the world works. People often argue that the power of data analytics lies in its ability to help us identify new patterns in the reams of data we now have access to. That’s true. But we should not forget that those new patterns often start as anomalies in the patterns we’ve already seen. Anomaly-driven innovation can help companies spot them and shape the nascent patterns forming around them.

The Power of Anomaly

Summary.

Partner Center