Analyzing the Dynamics of Shopping and Consumption: A Q&A with Professor Edward J. Fox

Many different kinds of disputes hinge on the choices consumers make, including matters pertaining to trademark infringement, false advertising, and drug pricing.

But the environment in which consumers make choices is changing due to an unprecedented combination of market, technological, environmental, and social forces. This changing environment can have a profound impact on certain purchase decisions that consumers make.

To explore the implications of these factors for consumer-focused litigation, Vice Presidents Laura O’Laughlin and Michael Carson talked with Analysis Group affiliate Edward J. Fox, the W.R. & Judy Howell Director of the JCPenney Center for Retail Excellence and Chair in Marketing at the SMU Cox School of Business.

Professor Fox’s technical training is in Bayesian econometrics (the application of statistics to economic problems). His research focuses on using customer-level data to identify the drivers of shopping behavior and make shopping and spending predictions. He also developed the data analytics core curriculum for all graduate business students at Southern Methodist University.

Ms. O’Laughlin: Why is it important to have a rigorous scientific basis for understanding how consumers make choices?

Professor Fox: We can interview consumers, but we can’t confidently generalize from what a few people say. We can also apply economic and psychological theories, but different theories offer different predictions about consumer behavior.

So, we gather data that include actual choices made in the relevant context in order to determine what consumers do in real life, not just what they should do or what we expect them to do.

The thing is, the planning, time, and cognitive effort invested in purchase decisions all affect how consumers process information and make those decisions. For example, consumers make decisions in different ways for different types of purchases. For cars or homes, decisions are usually made only after consumers have given the choice a lot of thought. On the other hand, decisions about which coffee or soft drink to buy are usually habitual. Most types of products fall somewhere in between.

Further, some purchases are planned while others are unplanned; the latter are often called “impulse purchases.”

It is also important to consider consumer choices in context. For example, when a consumer chooses a store to shop, she does so in anticipation of what she might buy there. More specifically, she might consider whether that store carries the products she is interested in, as well as promotions and prices on those and other products. This perspective on consumer choice can be particularly relevant to the calculation of damages.

So, empirical analysis enables us to learn, in context, how purchase decisions are actually made for different types of products or services, taking into account all these factors.

Mr. Carson: How do different methods work together to help you identify the characteristics or factors that drive consumer choice and behavior?

Professor Fox: If you use multiple methods to help answer questions about consumer choice, you can be confident that your answers don’t depend on exactly “how” the question was asked or the response was measured. If we measure a construct in multiple ways and the results are highly correlated, we have what is often referred to as “convergent validity,” which affords higher confidence in those results.

When addressing a research question, if we get very similar answers from multiple experiments or from a survey in which the answer is elicited in different ways, we have also validated our results. Similarly, if an experimental study and an observational study both have very similar findings, we have validated those findings and so have higher confidence in them.

Mr. Carson: When submitting expert reports or testifying, how do you integrate different kinds of evidence, such as survey evidence and sales data?

Professor Fox: In general, we gather evidence in two fundamentally different ways: by using data that were gathered in the course of doing business or for some other purpose; and by gathering data specifically to address our research question. The first type of data, commonly referred to as secondary data, may come from sales, costs, or other enterprise databases; it may also come from POS [point-of-sale] data or from syndicated data providers. This type of data is usually easier and faster to gather.

Unfortunately, secondary data can’t always effectively address our specific research question, and it can be difficult to identify causes and effects from secondary data. In contrast, the second type of data, commonly called primary data, can be tailored precisely to our research question. This kind of data can be gathered from scientifically constructed surveys or choice experiments. But for that reason, primary data are usually more time-consuming to gather and may be difficult to gather in a natural decision context.

Both types of data have benefits and drawbacks, but the combination of the two can be powerful. Together, they can often be used to address both the “what” and the “why,” creating a compelling explanation for findings while excluding alternative explanations. Furthermore, identifying the same effect in primary and secondary data confirms our findings, increasing our confidence in their accuracy.

Ms. O’Laughlin: What kinds of methodological tools or approaches could you use to analyze the primary data you collect on the choices consumers make?

Professor Fox: Consumer choice experiments can be particularly useful. Choice experiments allow us to apply a rigorous, quantitative approach to assessing how consumers make decisions. Conjoint analysis is such an approach – designing an attribute-based choice experiment and then applying a robust flexible modeling framework to analyze the responses.

How do we use choice experiments? One important application is inferring the importance of different “decision factors.” Which factors affect choices, how much, and for which consumers?

Another application is evaluating actual and hypothetical scenarios by predicting sales or market shares. For example, we can use choice experiments to forecast future sales or to estimate what sales would have been at a different price point or for a product with different characteristics.

These methods can be very useful in matters arising in litigation, from identifying and comparing the drivers of consumer purchase decisions to calculating sellers’ lost revenues and profits for damages.

Mr. Carson: The shopping environment had been moving more and more online even before COVID-19 hit, and people also have been relying more and more on their mobile devices. The pandemic only accelerated these trends. Does the growth in online shopping and the use of mobile devices change anything?

Professor Fox: Those trends have indeed changed many things. For starters, consumers have access to many more options for purchasing branded products. Historically, retail competition has been defined in geographic markets, but online shopping removes geographic boundaries and redefines those retail markets. It is likely that geographic and online retail markets will coexist for some time, offering consumers more varied options for shopping.

At the same time, markets have been created that bear only a modest resemblance to traditional markets. Search platforms, social media platforms, and even some retail marketplaces have characteristics that are fundamentally different from those that previously defined markets. For example, these new markets offer consumers access to a vast array of content and products, yet consumers may get that access for free. From the consumer’s point of view, these platforms seem to be offering something for nothing.

Of course, it’s not that simple. The “price” may not be in terms of dollars and cents, but in providing individualized data that allow targeted ads to be delivered to your device.

Online shopping – and, more generally, the pervasive use of mobile devices – also changes how consumers access information that drives their choices. Where consumers search, how they search, what information they access – each of these behaviors is changing, almost in real time.

Retailers’ point-of-purchase activities are also changing. Consider which seller an online retail marketplace selects as the default choice for a branded product. This selection of the “opt-out” option in a marketplace is valuable for a seller.

Another implication of mobile device use is how it impacts purchase planning. Shopping lists can now be maintained digitally and curated by the retailer or third-party provider. By using purchase histories to offer product recommendations and suggestions, firms can influence the consumer’s choices in new and different ways.

These kinds of developments raise a host of new and unanswered questions. What are these choices worth to the consumer or to the provider? How should they be valued? What are the competitive implications? Choice experiments and other methodologies will be critical in helping quantify value in this new purchase environment.

 

Ms. O’Laughlin: What types of data might you use to assess the levels of mobile and online shopping, as well as to better understand what factors drive consumer behavior in mobile or online shopping environments? How accessible are these data?

Professor Fox: Consumer decision making in the digital age is an important, rapidly evolving area of study. Perhaps it can be best understood by considering different types of data and how they may be applied.

One type of data is “aggregated” – over time, over consumers, over sellers, over geographies. Consider basic category-level metrics, for example, such as share of requirements online versus in-store. Consider also time-series data for online sales, digital device usage, clicks, cart abandonment, etc. We can use aggregated data to assess similarities and differences between product categories, sellers, and geographies, as well as how they change over time.

Perhaps more intriguing is “disaggregated” data, including individual-level clickstream, email, and text data. Augmented with information about individual behaviors (like purchases and visits to physical stores), disaggregated data can provide compelling insights about purchase drivers and the impact of sellers’ marketing activities, as well as customer similarities and differences.

We can analyze this type of data using “customer journey mapping,” a rapidly developing collection of tools for identifying the information consumers access along with whether, when, and how they progress to the point of purchase. Customer journey mapping has become increasingly important as consumers engage with the world digitally, a phenomenon that has accelerated during the pandemic.

At present, there is no generally accepted approach to customer journey mapping. But fortunately, it is amenable to rigorous quantitative analysis. Points along the customer journey can be analyzed separately – email click-through rates, affiliate link conversions, or shopping cart abandonments, for example. And the sequence of the customer journey can be analyzed using path modeling, structural equations modeling, and other sophisticated techniques.