In the following excerpt from their chapter “The Consumer INSIGHT Framework: A Hypothesis-Driven Approach to Data Analytics” in the forthcoming third edition of Kellogg on Marketing, Kellogg marketing professors Derek Rucker and Aparna Labroo use their extensive experience in research, teaching, and consulting to call for a shift in the way marketers approach data analytics. The shift involves moving from immediate action and toward hypothesis testing via their novel INSIGHT framework.
The Need for a Hypothesis-Generation Mindset
Rather than generate and test hypotheses in response to data, marketers often act with immediacy. As a result, one of marketing’s fundamental bedrocks—strategy—can languish and even risk being forgotten. The problem is that such behavior is analogous to pressing down the accelerator to speed forward in an automobile without a working steering wheel. Yes, you will get somewhere fast, but that somewhere might not be a desired destination at all. Let us share an example to better illustrate what is meant by an action-oriented as opposed to a process-oriented approach.
Consider a situation faced by Procter & Gamble and their Head & Shoulders shampoo in India. From 2014 to 2017, the brand faced decreasing market share despite media presence and distribution being stable. A tactical, reactionary solution might have involved increasing media spends and improving distribution. Perhaps P&G might have directed resources to markets where competitors were gaining share. Big data could have spotlighted the markets where share was declining fastest and where competitors gained the most. However, while this would have allowed immediate action, this solution could also have been suboptimal or even incorrect. Many factors could have caused share decline. The product may not have been effective, competitors may have had a more effective product, or consumers may not have had a need for such a product. Directing resources to media and distribution, while actionable, would not have solved any of these problems.
Instead of immediately acting, the team at P&G tested and ultimately dismissed these hypotheses. They offered an alternative hypothesis: Consumers believed that any shampoo would prevent dandruff. Rather than directing resources to increasing brand awareness or improving distribution, the team focused resources on solving the problem they had identified—that their consumers believed that any shampoo prevented dandruff. The brand ran advertisements that showcased embarrassing instances of having dandruff because other shampoos failed to prevent it. Instead of acting with immediacy, the brand team took time to develop and test hypotheses to understand the underlying cause of the problem and respond accordingly. We suggest that brands must embrace such a process orientation to successfully survive in the marketplace.
The Action-Oriented Approach
The action-oriented approach refers to the propensity to observe data and take an action in response to those data. This approach is necessarily backward-looking and reactive, not forward-looking and proactive. For example, a brand manager might receive data that suggest they are losing share in a market; in response, they might reduce prices, run a promotion, or try to copy the customer experience the competitor offers. But competitors get ahead when they do things based on their strengths; by following a stronger competitor, the success of the me-too tactician manager is doubtful.
Take, for instance, Art Henks. For two years after he became CEO at Gap Inc. in 2015, the company struggled with declining sales, the rise of highly successful fast-fashion competitors such as Zara that commoditized fashionwear through their reliance on big data and predictive algorithms, and the growth of e-commerce. As a result, Henks fired all creatives and, copying his competitors, turned his efforts toward big data instead.
As history revealed, simply turning toward big data and AI would not solve the problem. The following year, Gap’s profitability decreased even further. While Henks took immediate action, the action itself neither identified nor solved the true problem: that Gap Inc. had failed to update its iconic 1990s positioning of being a classic, all-American, understated, uber-cool brand that was a trend setter. By reacting to and imitating its major competitor, Zara, which had 10 times the revenue at the time as Gap, Henks possibly set up Gap for failure. Henks copied competitors’ strengths, at which Gap was not set up to excel. Unlike Gap, Zara was successful because its core business model was to commoditize fashion as a trend follower.
Every aspect from its retail and production to promotion and distribution was geared toward this differentiating advantage. With its factories located close to fashion capitals, Zara was able to identify trends early. By retailing in the plushest locations next to the most high-end fashionwear, it benefited from an aura of luxury and saved on advertising costs. By only producing in limited batches, it offered reasonably priced products but with strong margins. Most of all, in its reliance on big data—unlike Gap, which had a history of trying to push underselling products at discount—Zara had extensive records of successful products that would sell out because every aspect of its business was set up to ensure this success. Thus, unless Gap changed all aspects of its business to become exactly like Zara, it was unlikely to have chosen the right path to become successful. Moreover, even if Gap changed its entire business model to imitate Zara, it would not have the data pertaining to decades of successful sales growth. As such, it’s unclear whether reacting to its problems by copying Zara could ever lead to a differentiating advantage.
The Process-Oriented Approach
We propose the process-oriented approach as an alternative to the action-oriented approach. With this approach, managers use data to understand, test, and explain why a problem is occurring. Rather than react immediately to data, managers first generate alternative, and a broader range of, explanations for why something is being observed. A focus on explanations—the process underlying a problem or behavior—is central to the process-oriented approach.
Consider a situation where a competitor advertises on a new social media platform. When using the action-oriented approach, a brand manager could respond immediately by also advertising on the platform, increasing their advertising presence elsewhere, running price promotions, and so forth. In contrast, the process-oriented approach encourages a brand manager to consider what problem the competitors’ action introduces. That is, rather than reacting with action, the manager identifies what problem, if any, is introduced by the data. If a problem is identified—for example, the new platform suggests share may be stolen—the process-oriented approach asks the manager to generate hypotheses to explain why this problem has arisen. Why does advertising on this channel lead to a potential loss of share? Is it because the channel itself is central to consumers’ decision-making? Is it because the advertisements convey a new benefit that undermines the brand’s position? Generating hypotheses allows for discussion and testing, which offers insight to researchers on how to prioritize data.
In fact, hypothesis generation can reveal the importance of data that may not have been prominent or available when the problem first arose. Thus, we advocate that managers investigate their data to check which of their hypotheses may be the most viable, rather than allow themselves to be pushed by data that happen to be momentarily salient or react based on a sliver of data. If a new competitor is entering the market, a brand manager should ask whether, why, and how consumers might change their behavior in response to the competitor, and which consumers are more likely to react in the proposed manner. Although this argument might come across as simple in form, at its core it represents a philosophical shift from asking which marketing lever to pull in response to data (i.e., an action orientation) to focusing on the right data and understanding what the data say in response to the generated hypotheses (i.e., a process orientation).
To further illustrate this difference, consider a government that is concerned about the number of accidents its citizens have around trains. The action-oriented approach would likely lead the government to consider, for example, posting warning signs near train tracks, advertising the number of fatalities that occur every year, using social media channels to encourage train safety, or even changing safety regulations. These actions could be meaningful in changing year-over- year accidents and fatalities. However, the emphasis is on the levers that can be pulled to produce an effect of interest; taking these actions ultimately requires little insight, if any insight at all, into what causes accidents around trains in the first place.
In contrast, a process-oriented approach would go deeper by asking why people are having accidents around trains. Focusing on the explanation for the consumer behavior essentially calls for generating hypotheses about the process that explains why a behavior occurs. Rather than asking what it can do in response to accidents, the government asks why accidents or fatalities happen. Have previous efforts to educate people around train safety failed? Do people realize trains are dangerous? Do people enjoy danger? Notice that as we start to ask about the explanation or process, we have more enriched questions that can serve to generate hypotheses to inform our marketing efforts. Even if we want to pull a particular marketing lever—such as advertising—it is not the act of spending on advertising that reduces accidents. Advertising must influence behavior by solving a particular problem. And, as one begins to ask why advertising or any marketing activity exerts an effect, it becomes possible to understand which lever out of a potentially unlimited number of levers one must pull, and when one should pull that lever.
Implications of Process-Oriented Approaches
On its surface, our line of argumentation, although seemingly simplistic, represents a significant departure from current marketing practice. To conclude the story on train safety, an action-oriented approach may lead to the simple and relatively easy-to-implement conclusion that the government should advertise being safe, so it simply needs to think about what venues to use. In contrast, a process-oriented relationship would force the government to think about what is causing the problem and thus why and under what conditions an action such as advertising would be valuable. In essence, a process-oriented approach demands that practitioners obtain insight about the consumer.
In the case of train safety, one reason accidents occur is that people believe they know how to be safe; consequently, directly advertising to them about train safety would fall on deaf ears. While they might need to be educated, they do not want to listen to information they believe they already know. As a consequence, simply pulling the “advertising lever” would not be effective. Rather, it is necessary to figure out how to bypass the belief people have that they already know about train safety.
Although this example is illustrative, it has a real-world analog. In Australia, a public-safety campaign was run to encourage train safety. However, rather than just placing signs or advertising being safe, the advertisements discussed “Dumb Ways to Die.” The campaign featured a catchy song about numerous dumb ways to die; only at the end of several minutes did the campaign reveal that being stupid around trains is also a dumb way to die. The campaign waited to identify the true message so that consumers did not disengage because of content they believed they already knew. The campaign was reported to be associated with double-digit reductions in accidents after its launch.
The distinction between action-oriented and process-oriented approaches has a fundamental implication for how data should be used. In its strongest form, an action-oriented approach takes in data and reacts to those data.
Because an action-oriented approach prioritizes action, it sacrifices the prioritization of which data should be used. Thus, it may end up employing more data, using more irrelevant data, and responding in a sporadic and incoherent way to more salient data. In contrast, a process-oriented approach challenges managers to think about the underlying problem and thus prioritizes the use of data that informs that problem. Doing so helps managers sift through the data and reduces the quantity of data used, facilitates the use of more relevant data, and removes less relevant data. The result is that a process-oriented approach helps managers formulate a course of action that is hard-hitting, coherent, and incisive at hammering out the root cause of a problem.