Models of consumer preferences and choice most often follow a rational paradigm of decision-making — the consumer weights the values of attributes of each alternative considered  to derive its total utility value, then compares those utilities and chooses the alternative achieving maximum utility. Assuming this well-defined and straightforward process, it is relatively easy to construct predictions of preference shares or market shares. Yet the decision rule hereby described  is not so trivial or easy for consumers to apply; it requires a relatively large amount of information and a non-negligible computation effort. In order to simplify the decision process, consumers may use any of a selection of short-cut rules or heuristics (alternative-based and attribute-based rules) that demand less information and less cognitive effort to make a choice. On many occasions the process may not even include the rational-optimizing weighted additive (WADD) rule. Hence, decision processes tend to be more diversified than normally assumed in marketing research.

When consumers apply rules other than the WADD rule, predicted preference shares based on a maximum utility criterion for choice may off-shoot consumer actual choices. Indeed, there are other factors that may be implicated in biasing predicted preference shares when measuring preferences in consumer surveys — factors such as product availability and the visual display of products in stores, accessibility of brand and product information in the real world, and time constraints. Preference shares may further deviate from actual market shares when quantity is a key factor (e.g., a consumer buys 1, 2 or 4 cups ofyoghurt on any single purchase occasion and when purchase frequency varies). However, I will focus in this post-article on the factor that resides within the consumers, that is, the decision rules they use. In fact, decision rules are elected by consumers in adaptation to situational and environmental conditions as noted above.

Different techniques have been developed since the 1950s for identifying and tracking the rules people utilise to reach a choice decision between alternatives (e.g., travel destinations, car brand and model). Not less important than identifying single rules is the task of mapping the sequence in which these are used in a complete decision process, and this is where the study of decision-making can become complex.

• At the foundation of research in this field we find the “think-aloud” verbal protocol method for recording and mapping decision processes initiated by Simon and Newell. In this method, a consumer is given a choice problem and is requested to talk aloud  whatever thoughts come to his/her mind while (or right after) performing the task and reaching a decision (note: verbalising the thoughts aloud, not explaining them!) The content of protocols is later coded into procedures or mental operations that make up generic rules, and decision processes are mapped.
• A related technique uses information display boards of brands and their associated product attributes. By tracking the sequences in which consumers retrieve information items (i.e., cells in a brand-attribute matrix) the researcher can infer what rules are likely to have been applied and map decision processes.  This approach is implemented in a software known as MouseLab for computer-aided data collection and analysis of the decision processes (now also available as an Internet-based application MouselabWEB). Researchers Payne, Bettmann and Johnson who developed the original software applied it for their ground-breaking work on the behaviour of the adaptive decision maker in selecting rules and constructing decision strategies.
• Another interesting approach (Active Information Search [Brucks]) that allows decision makers to pose their own queries for product information (e.g., physical attributes, usage situations, price) may be seen as a semi-structured level between the protocol approach (least structured) and the approach of MouseLab (most structured).

More recently researchers in the area of decision-making have shown increased interest in the methodology of eye tracking for investigating search patterns for information. By measuring eye movements with specialised optical equipment, researchers capture locations where the eye fixates in a stimulus display (i.e., takes-in information) in-between rapid movements (saccades). This approach is already adopted often in studies of advertising effectiveness and shopper behaviour in retail scenes.

Researchers Reisen, Hoffrage, and Mast took the challenge of evaluating different techniques for tracking consumer decision processes (2008*). Primarily, they developed a multi-method framework approach called InterActive Process Tracing (IAPT). The investigative procedure has three stages:

1. Active Information Search (selecting relevant attributes);
2. Seeking specific attribute values of product alternatives (information acquisition) and choosing a preferred alternative;
3. Verbal reporting of decision processes (i.e., retrospective but with assistance from a moderator in formalising descriptions of the decision rules applied).

Firstly, the IAPT approach allows to integrate three methods, one at each stage, and compare their contributions to our understanding of decision processes and predicting choices. Secondly, the researchers provide new methodological insights by comparing two alternative methods for the second stage: the more veteran method of MouseLab and the new method of eye tracking for recording search patterns.

• In the second stage of information acquisition and choice, participants were shown several choice sets. Each choice set comprised a different selection of four alternative mobile phones (brands and their models) randomly drawn from a pool of phones available in the market. The form in which information on alternatives in a choice set is displayed is contingent on the method used for registering information acquisition: by mouse clicks or by eye fixations.

Let us consider first some key findings and insights obtained in the first study of Reisen and his colleagues, using MouseLab in the second stage for recording information acquisition:

• Two major types of strategies were identified: additive strategies (alternative-based) with equal or varied weights assigned to attributes or elimination (by-attribute) simplifying strategies — the elimination strategies were used more frequently (30 out of 31 respondents) than additive strategies (23 out of 31) though most participants (71%) combined strategies of both types.
• The difference in predictive ability of choices between additive (optimizing) strategies and elimination (simplifying) strategies is less dramatic as many might expect (55%-57% versus 47%-51%, respectively). Notably, predictive ability when using rules as described by participants in their individual protocols was much higher (73%) than in case of applying rules pre-defined according to literature as done for the comparison of strategies above.
• When using an elimination strategy, logic suggests that one would not seek more information on an alternative after it had been eliminated. Interestingly, however, acquisition of specific attribute values was traced for an alternative that was supposedly already eliminated by respondents using an elimination-by-attribute type of strategy (67%). This  makes sense, according to the researchers, if consumers initially acquired information as they explored the alternatives available for choice and only in a second phase committed to a decision strategy (i.e., the strategy eventually described in the verbal protocol).

In the second study, Reisen, Hoffrage, and Mast tested two methods for tracking search and information acquisition in the second stage: MouseLab and eye tracking, the more recent entry into this field of research. The information acquisition of each participant in this study was measured using both techniques sequentially (i.e., first series of choice tasks with MouseLab, and the second series with eye tracking [or vice versa]; half of the choice sets overlapped between series of choice tasks).  The following are noteworthy findings and insights:

• Frequencies of usage of additive and elimination strategies were similar in relative terms to those found in the first study.
• Predictive ability in choice tasks applying MouseLab was a little higher (69%) than in choice tasks with eye tracking (63%), but not statistically significant. More importantly, in repeated choice tasks (i.e., the same choice set), when respondents remained consistent in their choice under both conditions, predictive ability of the choice of that same alternative was considerably higher (78%) than in inconsistent tasks (40%).
• Respondents took more time overall to acquire information by mouse-click (MouseLab: 37 seconds) than by just eye gazing at the display (Eye Tracking: 20 seconds).
• However, eye gazing appears as more time-efficient from a respondent’s perspective: when eye tracking respondents accessed items in cells about 42 times on average compared with 22 cell-accesses while employing MouseLab. It is noted that the number of different cells accessed was similar between conditions (15-17 cells). Put differently, eye tracking allows for a higher rate of re-acquiring or double-checking information items rather than inspecting more of the information available on alternatives.
• In this study again it was observed that “regardless of the condition [MouseLab or Eye Tracking], participants accessed about 50% more information than prescribed by their strategy.” Yet, it is also reported that considering the information that is needed by a strategy, the greater part of it (82%) was accessed as expected. That is, the excessive information acquisition does not come necessarily at the expense of required information.

• Note: Findings from the second study suggest enlightening clues as to how respondents tackle choice tasks in choice-based conjoint studies. In particular, it points at the selective manner in which respondents consider information on alternatives included in a choice task, looking at only some of the attribute values shown, and yet they may re-access those same values several times until making their choice. Apparently this process may take less than 30 seconds to complete.

Following their assessment of findings in the second study, Reisen and his colleagues seem unconvinced that using eye tracking in IAPT is advantageous over MouseLab. With eye tracking respondents can finish each a choice task more quickly and yet access more information. However, they tend to repeatedly access the same information for a purpose the researchers describe as “validating a tentative choice.” Further difficulties in using the eye tracking methodology for tracing decision processes are suspected inaccuracies in capturing information acquisition (i.e., a respondent accidentally fixates on a cell while thinking; when respondents voluntarily mouse-click cells, the search paths seem more systematic) and failed calibration of equipment that leads to loss of unusable data.   The researchers conclude (p. 655):

It appears that this methodology improves neither the exactness of the description of the cognitive processes nor the quality of the results concerning the information search. Although this method allows for a more natural way of searching for information, it does not provide more informative data than does Mouselab.

Eye tracking can teach us a lot about how consumers look at and attend to different portions of ads such as bodies of pictures, text and brand logos, their  appreciation of package designs, or how shoppers inspect product displays on supermarket shelves. It can be helpful also in studying the decision processes consumers follow, but it is not more appropriate and accurate than former methods known for a similar purpose like MouseLab. More importantly, each of the four methods considered for IAPT specialises in capturing different aspects of the decision process (e.g., characterising patterns of information acquisitions vis-a-vis identifying the decision rule applied). A primary lesson to be taken from this research is that using multiple complementary methods with different scopes of specialisations can contribute considerably to obtaining a better mapping of decision processes and building models with higher predictive ability.

Ron Ventura, Ph.D. (Marketing)

Reference:

Identifying Decision Strategies in a Consumer Choice Situation, Nils Reisen, Ulrich Hoffrage, and Fred W. Mast, 2008, Judgment and Decision Making, 3 (8), pp. 641-658.