Editor's note: Based in the U.K., Theano Anastasopoulou
is a consultant methodologist at research company Kantar Health.

One frequently-used statistical technique within market research is choice-based conjoint (CBC). Leaving aside its mathematical foundations and just thinking about its application, CBC is a questioning technique that uses bundles of elements that describe a product or a service. These bundles are presented to consumers and, in most cases, choices and purchase intentions are obtained. In a simple conjoint task, we would, for example, ask for a choice between two mobile phone package bundles, one at £8 a month including 500 texts and 200 minutes and another at £12 a month including 500 texts and 400 minutes. We would present different combinations of those elements in a series of bundle pairs and would obtain a series of choices.

A systematic production and testing of the series of bundle pairs allows us to understand the influence of each of the elements that make up the bundles. For example, we can test whether people are willing to pay more for extra minutes and, if yes, what premium they are prepared to pay for different levels of increased minutes included in the package.

When conjoint results are produced, a typical output would be the utility value associated with different options of number of minutes included in a package. (These utilities are the typical conjoint output and are measures of how appealing or wanted a specific level of an offer is. They are used to estimate uptake at a later step but are not a measure of uptake.)

It will be no surprise to a conjoint analyst or user that, in the hypothetical example shown here, although the number of minutes increases by equal steps of 200, the corresponding utility increase is not symmetrical and equal in every step. We would happily accept this finding, considering that we expect consumers to evaluate the minutes offered against the context of their mobile phone use and their needs. So it is reasonable to expect that offering 200 rather than no minutes at all would make a package significantly more appealing, while moving from 400 to 600 minutes is not considered to be appealing to a similar degree. It is also relatively easy to explain why moving from 1,000 to 1,200 inclusive minutes does not appear as appealing because many consumers would consider 1,000 minutes enough to cover their needs. A further increase is not necessarily perceived as sufficiently beneficial.

The above findings provide valuable insight to product managers and marketers designing product or service portfolios as they allow them to uncover the perceived value and desirability of different elements of their offer. It is important the test covers the full range of options likely to be offered by a company in order to allow testing of all the possible alternatives and to establish their value. Another important recommendation is to repeat such an exercise at regular intervals, especially when there have been significant changes to the market. If, for example, most of the packages offer a minimum of 400 minutes, we would expect the loss of utility when moving from 400 minutes to a lower level to be higher than the increase in utility when moving from 400 to 600 minutes. If a year down the line the typical inclusive offer increases to 600 minutes we would expect a similar conjoint test to come up with different results. Consumers’ needs constantly change and utilities of different offers are expected to reflect those changes.

From a pure economics point of view

Now let’s move outside the market research mind-set for a moment and try to imagine that we approach this question from a pure economics point of view. Neoclassical economists using the “rational agent expected utility” theory tend to see people as rational agents that base their decisions on the objective value of an offer. If we tried to analyze the utility of each one of our inclusive minutes above rationally we would expect that the utility associated with increasing an offer by 200 minutes would be the same regardless of whether you move from 400 to 600 or from 200 to 400. In such a case, we could have tested the appeal of two levels having a 200-minute difference and extrapolate the utility change between them to all levels we want to examine. This would have saved us a lot of time and effort testing.

Any researcher would protest to such a suggestion as it would lead to inaccurate conclusions. It is interesting that something that may be unacceptable to researchers may have been perfectly acceptable to neoclassical economists. On the other hand, psychologists studying human choices would have sided with the market researchers.

Different disciplines may be working independently, completely unaware of practices and learnings in other disciplines. Psychologists studying human decision-making have been happily working independently of economists who were developing and using theories about economic behavior for years and never contributed or inputted to this development with their psychological behavioral insights. Then a team of psychologists, upon coming across the rational agent expected utility theory in economics, were surprised by how it conflicted with the sometimes irrational human behavior they observed in their research. These researchers decided to spend some time challenging the rational economic theory and put forward an alternative.

One of those psychologists was Daniel Kahneman, who in 2002 was awarded a Nobel prize for his contribution to the field of economics; this contribution mainly originated from his influential work on prospect theory, which was the response to economic models of the time (his collaborator on this theory, Amos Tversky, unfortunately passed away six years before the Nobel was awarded). The significance of this Nobel is indicated by the fact that the only other psychologist awarded a Nobel was Pavlov (and he may even be considered as a biologist).

A more accurate explanation

Although prospect theory applies to a very specific task, what Kahneman and Tversky achieved was to offer a more accurate explanation of human decision-making that takes into account certain cognitive features that underlie human decision making. The main contribution by their theory was the introduction of a reference point that humans use and is very important when coming to a decision; this reference point is determined by current circumstances or the context for their decisions. Their work was groundbreaking as it demonstrated in a powerful way how psychology can inform the field of economics. This was one of the key milestones in the development of behavioral economics, seen by some as a revolution in economic theory as it has since brought other fields like psychology, sociology and political science into the field of economics.

Prospect theory is mainly based on three key cognitive features. As mentioned earlier, evaluation against a reference point is one of the key cognitive features. In our conjoint example, a reference point that underlies a decision is that a consumer shopping for a mobile package expects at least some minutes to be included in the package rather than none, so it is not surprising that offering fewer minutes than the reference point or none at all will make the package unappealing. Another key cognitive feature described by prospect theory is loss aversion. If most packages offer 400 minutes, then the majority of consumers are likely to have this in their current package. Therefore, a reduction from this level is likely to lead to a larger utility reduction than an increase from this level of a similar magnitude. Finally, another key cognitive principle is this of diminishing returns. An increase of 200 minutes from a starting point of 1,000 is less attractive than an increase from a starting point of 400.

Design tests reflecting current expectations

The same principles will apply to testing of any product attribute. When we design product profiles in any sector (whether it is consumer products, business-to-business services or health care therapies) it is important to identify current expectations or accepted norms and then vary our options around this norm (also considering that these may differ by subgroup). It is important to test the effect of realistic levels of increases as well as decreases versus these norms and make sure we design tests reflecting current expectations.

Finally, new product launches that are likely to significantly move current benchmarks will need a different questioning and modelling approach: a different questioning approach to ensure the new reality is clearly de-scribed and reflected by our respondents before they give us their preferences and that we are realistic in how many and which variations of the future competitive landscape we can reliably test; and a different modelling approach to ensure we apply known cognitive principles to our analysis and extrapolations.