Editor's note: Michael Curtis is a research analyst with the Detroit automotive group of Maritz Marketing Research.

The importance of understanding major purchase decisions at the "natural group" level was first discussed in Quirk's Marketing Research Review in the article "Natural Group Interviewing" by David Pagnucco and Robert Quinn, which appeared in the December, 1988 issue. In the interim, Maritz Marketing Research has conducted some preliminary research among natural groups. The purpose of this article is to update the original article and to chart some future areas of study.

As Pagnucco and Quinn pointed out, research that focuses on product purchase behavior is generally conducted at the individual level. This is partly a result of the methodologies available, such as quantitative surveys and one-on-one qualitative interviews. These methods focus on individual attitudes and behaviors. The methods are appropriate in most instances, since purchase behavior is typically an individual phenomenon. There are purchase decisions, however, that cannot universally be assumed to occur at the individual level.

Groups of two or more individual oftentimes play a role in the decision making process. The decision making roles within a group may be implicit, as when one individual orders a pizza with specific toppings, taking into account the known preferences of other family members who will be eating it. The roles may also be explicitly played out, with group members taking a more active role in the purchasing process. This would seem to be most likely to occur with items that are highly salient to several members of the group. High cost, high involvement items such as housing, transportation and travel fall into this category. Natural group interviewing, then, is an approach designed to take into account the explicit interactions that occur within primary group decision making processes.

The Automotive Research Group of Maritz has included the natural group interviewing (NGI) concept as part of three major product clinics conducted between dune, 1989 and August, 1990. In June, 1989, 17 natural groups, consisting of husband and wife, were recruited to participate in an automotive clinic as an adjunct to the more typically recruited individual car owners. The groups were randomly divided into three clusters and different data collection strategies were applied to each.

Cluster One used a single questionnaire that was jointly completed by the couple. They were instructed to discuss each question before coming to a joint decision, then to record a single answer on the questionnaire. Cluster Two couples were given separate questionnaires to fill out and instructed not to discuss their answers. Cluster Three was similar to Cluster Two, except that following the completion of the questionnaire, the couple was instructed to sit down and jointly fill out a combined questionnaire based on input from their individual questionnaires. Following the quantitative data collection a short one-on-one interview was conducted with each couple to assess reaction the overall process. The intended purpose of this study was to gain some initial qualitative insight into the NGI process.

The second and third studies to make use of NGI were part of a product clinic conducted in Chicago and San Jose during August, 1990. The NGI process was used in a more mainstream fashion here. Recruiting for all clinic respondents was based on vehicle ownership and purchase decision makers for the vehicle. If more than one person was involved in the purchase decision of a vehicle the second decision maker was invited to participate. In no case was a third person invited as part of the same group. Most groups consisted of husband and wife, but in one case the group was made up of a father and his son.

Groups were screened to determine who the principal driver of each vehicle was. This person was assigned primary responsibility for completing the quantitative portion of the questionnaire. Both respondents were allowed to participate in the qualitative interviews that took place; similar in nature to a semi-structured focus group format. Upon completion of the study the interviewers were debriefed to obtain their perceptions of the NGI process.

The following perceptions are based on the three trials of NGI which were conducted.

Natural group decision making process

A significant subset of vehicle purchase decisions are apparently made by natural groups. Based on respondents who qualified and agreed to participate in two of our automotive clinics, 31% in Chicago and 40% in San Jose were natural groups. According to a 1987 Newsweek/Maritz Marketing Research study, 48% of new car purchase decisions are made by two or more people.

The overwhelming majority of groups that we interviewed, consisting primarily of married couples, indicated that individual members had equal involvement in their most recent new car purchase decision.

Natural groups use a number of approaches to resolve differences of opinion regarding vehicle purchase decisions. Some of the most common methods include:

  • Discuss differences of opinion, then reevaluate the vehicle.
  • Look at other attributes such as price, standard features, roominess, etc., to try to resolve the difference.
  • Find an alternative vehicle as a compromise.
  • Obtain more information with which to make a decision, such as through a test drive of the vehicle.
  • Use dealer options or after-market purchases to resolve differences.
  • Evaluate the importance of the feature or characteristic in dispute.
  • Set minimum vehicle standards and designate other attributes as flexible.
  • Give the primary driver final decision making power.
  • Have each member make decisions within their areas of expertise.

Quantitative data collection

Both data collection and interpretation are problematic. If each group member individually completes a questionnaire, how are the data to be combined in a meaningful fashion? Treating them as individual respondents biases the overall sample by weighting the data in favor of groups. Taking a group average for each question assumes equal input into each facet of the evaluation process-an assumption which may not be valid.

An alternative method seems to be a combined response where group members conduct their own form of "averaging." That is, the group completes the product evaluation in a manner they deem most appropriate. This format has been criticized because in some cases participation is very unequal. A husband may, for example, take charge of the evaluation, thereby allowing his wife very little input. An argument could be made, however, that the internal dynamics reflected in this situation would also be reflected in a purchase decision as well. The method becomes self weighting in this respect, with each group deciding internally whose opinions are most important on any individual question. Most of the groups seem to operate in this fashion, with an individual dominating a portion of the evaluation process. For example, the wife may be the primary evaluator of the interior since she will drive the vehicle most often. Her husband may have more mechanical knowledge and therefore his opinion will weigh more heavily with respect to powertrain issues. The end result using this method would seem to reflect most accurately the dynamics involved in the group's real-life purchase decision.

This method works well for product feature evaluations, but it is not so obviously correct for conducting image assessments. What happens when individual group members have very different images of a vehicle? Can a "natural group image" be developed that is consistent with the methodology used to collect image assessments from an individual? Is it more appropriate in such a case to collect individual responses from each group member? Perhaps assigning the task to the principal driver or decision maker provides the optimal solution; however, it begs the question of what role image plays in the group decision making process compared to individual decisions.

Other difficulties arise with the collection of demographic and psychographic data. Some factors, such as household income, are captured in the same way as always, and are thus unaffected. Age can be more of a problem to define, particularly when the natural group spans multiple generations. As suggested, perhaps the demographic and psychographic profiles should be obtained only from the principal driver or decision maker. On the other hand, perhaps individual profiles are meaningless in group purchase situations and need not be collected.

Qualitative data collection

A particularly effective use of natural group interviewing involves the collection of qualitative data. Using an interviewer in a semi-structured one-on-two interview (similar to a focus group format) can provide a gold mine of information. Individuals feed off of one another in providing insight into feature preferences and purchase decisions. This format is particularly useful when used to provide breadth and depth to quantitative data. This is also a particularly good avenue for exploring the self-weighted group dynamics as they impact the decision making process.

Show rates

Another factor we were interested in looking at was respondent show rates. Typically, anywhere from 60-90% of people recruited to participate in an automotive clinic actually show up. Rates tend to vary by city, location within the city, weather conditions, competing activities in the city and numerous other factors. Our first bout with recruiting natural groups entailed a small subsample of 17 groups in Chicago as part of a larger overall study. All 17 groups showed, for a 100% show rate.

In two subsequent studies, one in Chicago and another in San Jose, direct comparisons were made between the show rates of individuals and natural groups. In both cases, although sample sizes were small, the individual show rates were slightly higher. In Chicago, 64 of 78 individuals (82%) versus 26 of 37 natural groups (70%) showed up to participate in the study. The San Jose numbers were 50 of 60 (83%) and 30 of 40 (75%) respectively. Although neither result is statistically significant at a 95% confidence level, there is clearly no evidence to suggest that natural groups would provide a higher show rate, which is what we had originally hoped for.

Recruiting

Another area that natural groups may differ from individual respondents is in the initial recruiting stage. Potential respondents are contacted via phone and interviewed on a series of screening questions appropriate for the particular study being conducted. If the person answers all screening questions properly, he or she is invited to participate in the study. Some accept and others decline. Standard procedure has always been to allow only one individual per household to view the vehicles on display due to the high security nature of some of the pre-production vehicles. This was often disincentive for individuals since they were unable to bring another family member along. We felt that by integrating the natural group concept a higher proportion of acceptors would be obtained. Although no experimental data were obtained to directly compare acceptance rates, natural groups were more likely to agree to participate in the subjective view of the telephone interviewers who conducted the recruiting.

So where do we go from here?

Given the proportion of new car purchase decisions made by natural groups, it is imperative that the behavior be reflected in marketing research methodological designs. Problems discussed in this article with data collection at the group level should be addressed rather than ignored. Data collection may be cleaner by utilizing the standard, individual respondent approach, but erroneous results are just as likely to occur as any time you have a sample which is not representative of the targeted universe.

Using the group interview design, data collection problems occur. Addressing the measurement of traditional demographic, psychographic and image assessment variables becomes an important consideration. Is it possible, and more importantly, is it valid to obtain a group image assessment or group psychographic profile? The answer may be yes for groups such as married couples and no for multi-generational groups like father and son. Perhaps the problem can be alleviated by rethinking market segmentation strategies along a different series of variables.

Thus far, the groups we have addressed have been limited to two individuals; typically married couples. Is it worth the additional complexity to invite a third, or even fourth, decision maker when appropriate? Not only might it be useful, within certain vehicle segments at least, it may be highly desirable. For example, vehicles such as minivans may have very diverse usage patterns within the same household. Individual decision makers within the group may key on different features such as gas mileage, cargo capacity, people capacity, towing capabilities, maneuverability, etc. By not inviting all decision makers, one or more key features related to the actual purchase decision may be artificially downplayed.

This article raises more questions than it answers. Perhaps the most compelling one concerns whether natural group decisions can be accurately measured indirectly through an individual member. If so, the discussion put forth in the article can be largely ignored. If not, these issues must be addressed. An a priori argument would suggest that the burden of proof lies with proponents of the "individual as spokesperson" perspective.