Product configuration: a research technique for the times | Articles

Abstract

With more and more opportunities available for consumers to design their own products, a research tool that uses a similar approach makes perfect sense, the authors say.

Made to order

Editor's note: Rajan Sambandam is chief research officer at TRC, a Fort Washington, Pa., research firm. Pankaj Kumar is managing director of Quantelligence, the marketing analytics division of TRC.

Research produces the best results when it is most realistic. When consumers are made to go through the same decision-making process that they would have used for actual purchases, the information they provide tends to be more reliable and useful.

The trick is to be able to derive sufficient usable information from those choices in order to form judgments about decision-making that can help companies in their marketing efforts. Methods that have gained in popularity through the last couple of decades (such as discrete choice conjoint analysis) have done this quite successfully. They have used advanced statistical analysis that is still increasing in its complexity and ability to draw ever deeper insights. Compared to an era when preference information was gathered primarily through ratings scales and other context-free approaches, these choice-based methods had a significant advantage in terms of their realism.

Reality has shifted

But the reality of the marketplace has also shifted in the last decade, thanks to the widespread use of Internet commerce. Among its many significant advantages is its ability to allow consumers to become, effectively, producers. Think of how you buy a computer today. It’s not the manufacturer showing you fully-formed products to choose from (a process that a technique like discrete choice conjoint analysis imitates), it’s you telling the manufacturer what you want included in your product. In a sense you are building (or configuring) the product.

This process started out with products like computers that are more suited for this kind of consumer-driven building but is spreading to other areas, often quite unrelated. Now you can build your own clothes, shoes, credit cards, digital and entertainment services, insurance products, industrial equipment – the list goes on. And it will keep going on because the Internet is the central reality of our lives today. So, if consumers are increasingly inhabiting a world where they get to build their products, does it not make sense that research would be more realistic if it imitated that process to get better-quality information?

The answer is yes, of course. Because not only are we then keeping up with the reality of the marketplace, it also turns out that getting consumers to build their own products has other significant advantages. They will find the task more realistic and enjoyable – primarily because, unlike an approach like discrete choice conjoint, configuration is significantly less repetitive and tedious. By metaphorically looking over consumers’ shoulders as they make choices, we are able to glean interesting information on how they make decisions. Conjoint analysis is generally what is called a decompositional approach. People tell you something about the whole product and you then derive what is important to them. Building your own product is a compositional approach because respondents to a study have to make explicit decisions at each stage and build their product one piece at a time. This latter approach increasingly makes sense in a world where people are in fact building their own products.

Getting consumers to build products is an idea that has been around for several years. Eric Von Hippel of MIT has been talking about it for the last couple of decades, waiting for technology to catch up to his thinking. The Journal of Marketing Research published a seminal article a decade ago on consumers making choices and how that data can be analyzed. Menu-based conjoint as a practical approach is increasingly becoming a reality. In this context, having consumers configure complex products and analyzing the data to get useful conjoint-like information should be of interest to managers in diverse industries.

Starts with the design phase

So, what actually happens in a configuration study? It starts with the all-important design phase, where an in-depth discussion of the product or service in question is needed. Every feature and option that a company wants to offer and that consumers may desire has to be listed out for potential inclusion in the study. If you don’t ask, they can’t tell you anything about it. The next and perhaps even more important step is the imposition of constraints. If you were building your laptop on a manufacturer’s Web site and you knew you were going to get it for free, wouldn’t you build your dream product with little regard for the real world? The same thing will happen in the research if you don’t impose price (or other relevant) constraints on every option presented to respondents. In some industries this is straightforward. In others (such as insurance products) input from other internal departments is required to make numbers as accurate as possible. This is actually a useful situation as it gets more stakeholders involved with the study and makes them more receptive to the results.

There are a few more steps to finishing the design but these are the essential ones. Note that unlike in a discrete choice conjoint type of study there is no statistical underpinning to these designs. Such a feature is both an advantage of those methods (because of the efficiency) and a disadvantage (because of the reduced flexibility). A configuration study design has complexity, of course, but it is of a different variety and the flexibility of the design phase can be an order of magnitude higher. So, products with numerous complex interconnections (if color is red, shape cannot be round; if the plan is X then Feature Y cannot have Price Z, etc.) can be easily handled.

Programming creativity allows for an interesting and interactive experience for the respondent. It is easy to see why this kind of research is more enjoyable for respondents. At each stage the respondent picks the option to include and clearly-displayed price information makes the choice realistic. Running totals and options to make changes to the product allow respondents to design a product that fits the budget.

Results can be as simple as percentages of people who picked each option. This information is very useful and translates easily for managerial decision-making. Loads of additional information is available about the decisions people made at each step (and what that could mean), the most sought after options, the most frequent trade-ups and trade-downs – all in an easy-to-understand format. Respondents are effectively segmenting themselves based on their preferences but one could certainly segment them further in different ways. As mentioned before, more advanced analyses are also possible, leading to conjoint-type utilities and all their attendant advantages.

Auto insurance example

We’ll use an auto insurance example to illustrate. The topic is interesting for a few different reasons. All drivers need it and most adults choose their own provider. It can be customized for a driver and it has some complexity built into the process, especially with regard to differential pricing. The decision-making may not be straightforward, with rules being used to arrive at an optimal product. It is often renewed every six months, providing an opportunity to revisit the decision-making process with somewhat high frequency. And, of course, it is quite suitable for a configuration exercise.

The study was set up as a task for choosing auto insurance for oneself. A basic product (largely hewing to state-mandated minimums) was described followed by the configuration exercise where respondents were offered choices on six features. Each feature had three to four options including a base option and respondents could choose to stay with the base product or shift to one of the other offered options. Some options would increase the total price while others would lead to a lower price. Given the customized pricing used in auto insurance, we kept the task realistic by asking respondents for their current expenditure and using that as a basis for building the price for the overall product. Respondents build their ideal product from the choices provided as they proceed through the exercise.

Basic results

Of the 822 respondents in the study only 20 percent chose the base option in every feature. In other words, the vast majority of study respondents chose to alter the base product to fit their specifications, showing both their inclination with regard to auto insurance and their level of engagement with the exercise. In every feature there are considerable proportions of people choosing to upgrade (and sometimes trade down) from the base product. For example, as shown in Table 1, with collision deductible, a fifth of respondents show a willingness to lower it down to $250 even though it adds $125 to the overall premium. Another fifth would rather lower their premium by choosing higher levels ($1,000 or $2,000) of collision deductibles.

Almost half of respondents opt for some form of accident forgiveness option while about that proportion indicate they would prefer policy terms longer than six months. In both cases respondents are showing that they are willing to pay for such amenities, thus providing an auto insurance company with valuable input on pricing these kinds of innovative features. Profiling people by the choices they make also provides interesting information. This is clarified more when we run a segmentation analysis on the choices that people make when building the product. Using a neural network-based segmentation method (called self-organizing maps) we can identify segments with clearly distinguishing characteristics:

Segments are differentiated mainly based on deductible preferences.
A somewhat smaller high-deductible segment of consumers who are relatively affluent, educated and younger. They are much more interested in unusual offers like very high deductibles and more likely to indicate a willingness to buy the product that they have built.
A low-deductible segment that is older, almost as affluent, has more children, generally prefers dealing with an agent and doesn’t use the Internet to shop as much. Unusually low deductibles are especially attractive to this group, perhaps because they are more risk-averse than other segments.
There are also segments which tend to go with the base product offer and don’t show much inclination to customize the product. They do have some clear differences among them in terms of variables like accident forgiveness but it is clear that they are quite different from those who seek high or low deductibles.

Advanced modeling

The primary information that comes from a configuration exercise is simple, intuitive and very useful. But we don’t have to stop there. Advanced econometric modeling can be applied to the data to draw out conjoint-like insights even though the design is not set up accordingly. While the problem is quite complex because of the design flexibility, it is possible to derive individual-level utilities or attractiveness scores for every option in every feature using hierarchical Bayesian estimation and we were able to do that in this case. Of course, this provides the same level of flexibility on the back end that has been the hallmark of conjoint designs. In essence, we overcome the front-end design constraints of conjoint while availing ourselves of its back-end flexibility.

In our experience across several industries, it is possible but harder to generate individual-level utilities with these data compared to conjoint data. The reason is simple: Conjoint analysis uses an experimental design upfront to ensure that all options are viewed an appropriate number of times, while the input in this case is much more chaotic as it depends on the whims of individual respondents. However, our experience is that when validation tasks are included it is possible to obtain hit rates that are generally comparable to what can be seen in conjoint designs. Of course, a proper one-to-one comparison may not be possible as each method is usually appropriate for different problems and the design discipline of conjoint is always going to be an advantage. But given the design flexibility enjoyed by the configuration exercise, the ability to get good-quality utilities (along with all the other information) makes it a good way to tackle problems that can be too complex for other approaches.

So what does it practically mean to have individual-level utilities? Since utilities express the desirability of every option for every respondent, we can make calculations like take-rates of any product we choose to design or preference shares of any groups of products. We don’t have to restrict ourselves to only products that were actually designed by respondents. Since individual-level preferences are now available, any product combination from the universe of possibilities (often running into tens of thousands of products) can be created.

For example, if a high-deductible product and a low-deductible product with varying levels of accident forgiveness and policy terms were introduced, how would the market react to it? Which demographic groups would be more likely to choose one over the other? Who are the people who prefer high bodily injury liabilities? All these kinds of questions can be answered because we now know preferences at the individual respondent level. In fact, we can build a custom simulator that would allow all these scenarios to be played out to get a full understanding of how consumers make choices in the market place.

For simplicity, this example uses only six features and three to four levels per feature. In reality, the configuration exercise can handle far more features and levels, bounded only by respondent engagement and the ability to develop price constraints. And of course, this method can be applied in a variety of industries.

Likely to resonate

Product configuration is a deceptively simple and engaging way of gathering information from consumers by having them build their ideal product. In the process of building they provide considerable insight into their preferences, allowing companies to design products that are much more likely to resonate in the marketplace. In specific situations this method has the potential to surpass existing methods of preference elicitation (such as discrete choice conjoint) while at the same time providing an engaging and enjoyable experience for the respondent.