Correlates of Search Patterns For an Innovation

James W. Harvey, George Mason University
ABSTRACT - This study addresses several of the shortcomings of diffusion research as summarized by Rogers (1976, 1983) and Robertson (1984). To better understand the role of marketer-dominated, sociometric, and non-commercial sources of information in the adoption process, factor analysis was used to infer subJects' information search typology from an information display board. Using indexes reflecting source structure, subjects were grouped by cluster analysis into patterns of typology use. Zero-order correlations provide insight into demographic, task environment, and behavioral associations with types of search for a discontinuous innovation
[ to cite ]:
James W. Harvey (1986) ,"Correlates of Search Patterns For an Innovation", in NA - Advances in Consumer Research Volume 13, eds. Richard J. Lutz, Provo, UT : Association for Consumer Research, Pages: 414-418.

Advances in Consumer Research Volume 13, 1986      Pages 414-418

CORRELATES OF SEARCH PATTERNS FOR AN INNOVATION

James W. Harvey, George Mason University

ABSTRACT -

This study addresses several of the shortcomings of diffusion research as summarized by Rogers (1976, 1983) and Robertson (1984). To better understand the role of marketer-dominated, sociometric, and non-commercial sources of information in the adoption process, factor analysis was used to infer subJects' information search typology from an information display board. Using indexes reflecting source structure, subjects were grouped by cluster analysis into patterns of typology use. Zero-order correlations provide insight into demographic, task environment, and behavioral associations with types of search for a discontinuous innovation

OVERVIEW

Rogers (1976, 1983) and Robertson (1984) review several conceptual and research deficiencies in adoption theory which represent important challenges for the continued examination of this unique aspect of consumer behavior. These researchers argue that most innovation studies (1) neglect to use a process approach which tracks multiple dependent measures over time, and rely too heavily on retrospective measures; (2) focus on innovations of which adoption is assumed to be desirable; (3) fail to include thorough sociometric analysis, including multi-influence decision making; (4) avoid examination of innovations which are only tangentially important and are not highly recommended, for which low-involvement models of decision behavior may be most appropriate; (5) fail to consider that situations may exist where adopter categorization and innovator identification may not be relevant; (6) need a stronger emphasis on market segmentation; (7) avoid the study of innovations where performance of the offering is difficult to judge; and (8) neglect to account for the very real world influence of marketer-controlled communications such as advertising, personal selling, sales promotion, and packaging.

One aspect of the adoption process which draws together several of these seemingly disparate conceptual and research design issues is the magnitude and composition of external search. Search composition is determined by the information sought from various sources and has long been reported to have different effects on the prospective adopter. For example, commercial-source information detailing design, function and performance is thought to stimulate awareness and interest in the innovation, while support from professional and personal sources are needed to encourage trial (Katz 1961; Lionberger 1960; Ryan and Gross 1943).

However, most of the recent studies of search magnitude and composition (e.g., Newman and Staelin 1972), as well as those which examine effects of availability, format and number of alternatives (e.g., Jacoby 1984) and individual difference effects such as demographics, personality, cognitive structure and experience (e.g., Duncan and Olshavsky 1982; Schaninger and Sciglimpaglia 1980, 1981) base their findings on offerings other than innovations (for exceptions, see Arndt 1967; Berning and Jacoby 1974; Bettman 1970; Black 1983; Dickerson and Gentry 1983; Furse, Punj and Stewart 1984; Shoemaker and Shoaf 1975; Wilton and Pessemier 1981). Those studies which were conducted in an adoption of innovation context, however, possess many of the characteristics criticized by Rogers (1976, 1983) and Robertson (1984), outlined above. Consequently, more needs to be known about external search for innovations, using research designs which address these criticism.

RESEARCH GOALS

To address several of these shortcomings, six goals guided the development of this study: (1) to report the magnitude and composition of consumers' external search for an innovation in a real-time environment which would lessen the reliance on retrospective measures and strengthen the opportunity for both control and insight into process; (2) to make available to prospective adopters all possible combinations of commercial, non-commercial, personal and impersonal types of information; (3) to validate decision makers' perceptions of differences in these information sources; (4) to establish search patterns across information sources; (5) to assess the relations between information source and selected demographic and task environment measures, as well as behavior; and (6) to present these findings where the object of the study is a relatively inexpensive, comparatively unimportant product, true of most innovations to be found in the consumer marketplace.

METHOD

One hundred twenty-four women responsible for food preparation were recruited through a variety of civic, special interest, and church groups. Their average age was just over 35 (s = 14), with 1.8 children (s = 1.3), middle income and house value, with 55% college educated.

Upon arrival at their meeting room, subjects were seated away from the information display boards and handed a written product introduction which was described as a "press release to familiarize the public with a new texturized vegetable protein food product which was soon to become available in local supermarkets." Having read the press release, each subject was seated in front of an information display board (IDB) and told that the next thing requested of them was to select as much or little information beyond the press release as they would need to decide whether or not to try the product.

Each IDB contained sixteen cards which were visible to the subject: Package Front, Comment of a Family Member Who has Tasted Product, Sales Representative, Manufacturer's Reputation, Government Report: Flavoring, Product Ingredients, Consumer Reports Study, Print-Ad, Nutritional Information, Good Housekeeping Test, Cooking Instructions, Government Report: Coloring, Government Report: TVP, Comment of a Friend Who has Tasted, No Trial, and Trial. The women were told that the information labeled on the front of each card was contained on the back and that to obtain the desired information they were to remove the card from the hook, turn it over, read it, stack it in pile, and then choose whatever information was next desired. Subjects were told to repeat this process until "you have enough information to decide whether you want to try the product or not, and then select the card which reflects your choice: Trial or No Trial."

Fourteen of these pieces of information (Trial/No Trial Deleted) represent combinations of commercial, non-commercial, personal and impersonal sources, as outlined by Robertson, Zielinski and Ward (1984, p. 89) and serve as the primary dependent variables of the study. Figure 1 presents the categorization of these cues.

After completion of the search task, subjects completed a questionnaire which gave them an opportunity to "order" units of the product at $1.00 each, and contained the demographic and task environment variables which serve as the measures for this study. The homemakers were paid $8.00 for their time and cooperation.

FIGURE 1

SOURCE TYPOLOGY OF AVAILABLE CUES

RESULTS

Data Reduction

To infer the underlying structural character of information usage, counts were made on whether each of the fourteen cues was selected by the subjects. As advocated by Schaninger and Sciglimpaglia (1980), Rao's canonical factoring was applied to these fourteen types of information to infer subjects' taxonomy of information source. Rao's factoring with varimax rotation is appropriate in instances characterized by scale-free variables and where the elimination of multicollinearity is desired. Four factors with eigenvalues greater than 1 were found, explaining 59% of the original variance. Table 1 presents the factor pattern matrix, with decimals and loadings less than .3 omitted.

TABLE 1

FACTOR ANALYSIS OF THE SEARCH VARIABLES

With some exceptions which will be discussed later, the results of factor analysis appear to have considerable correspondence with the expected typology. Factor 1 (31% of the variance) can be labeled Nonmarketer Controlled/ Impersonal, with information from government, independent, and non-manufacturer testing sources being sought. Notice that Nutritional Panel has mixed loadings with Factor 2 and that it as well as Consumer Reports and Good House- keeping have comparatively low loadings on Factor 1. The level of these three loadings suggests that these cues do not fit well into this typology. The format and contents of Nutritional Panel may have been seen as somewhat objective and "scientific" despite being from a commercial source (albeit non-manufacturer), while Good Housekeeping and Consumer Reports may have been viewed as "nonmarketer," but not as much as the government reports.

Factor 2 (12%) is a Marketer Controlled/Impersonal factor, as seen by search from Package Front, Product Ingredients and Cooking Instructions. Note again the Nutritional Panel mixed loadings with Factor 1 and those of Print-Ad with Factor 3. The mixed loadings of Nutritional Panel across Factors 1 and 2 reaffirms the argument that this cue is less clearly seen as either marketer controlled or nonmarketer controlled, while the mixed loadings of Print-Ad make interpretation of Factor 3 problematic.

Factor 3 (8% of the variance) is a Marketer Controlled factor with both Personal (Sales Representative) and Impersonal (Print-Ad) information cues having high loadings. The low loadings of Manufacturer's Reputation reflects further lack of subject clarity as to how to classify this cue. Factor 3 may be best labeled, therefore, as a Marketer controlled-Advocate factor. Last, Factor 4 (8%) offers clearer interpretation. The high loadings of Family Comment and Friend's Comment support the conclusion that Factor 4 is a Nonmarketer controlled/Personal source of information.

The results of this analysis reasonably well confirm the proposition that consumers view information across two dimensions: marketer and nonmarketer controlled and personal and impersonal. What remains to be examined are the ways in which these factors are combined into patterns of use while subjects learned about the foot innovation in the study.

Search Patterns

To examine search patterns, the factor analysis described above was used to guide a follow-up cluster analysis in a fashion similar to that taken by Kiel and Layton (1981). Counts of the information cues comprising each of the four factors determined by the analysis presented in Table 1 were calculated for each subject and divided by the number of cues identified in each factor to lessen the effect of different numbers of cues comprising each factor (Nutritional Panel was assigned to Factor 1 and Print-Ad was assigned to Factor 3). Therefore, four new variables (indexes) were created for each subject which represented the percentage of information selected, represented by each of the four factors, ranging from 0 to 100. These search indexes were used as the input to K-Means clustering of cases, BMDPKM (Dixon 1983). Euclidean distances of the unstandardized data was used as the grouping criterion. A five cluster solution was chosen based on spreads of centroids, cluster size and parsimony. The F-ratios for the four search indexes within each cluster, ranged from 13 to 133 (df=4,119;p=.000). Pooled within cluster correlations ranged from |.03|to |.17|. A summary of this analysis is reported in Table 2.

TABLE 2

CLUSTER ANALYSIS RESULTS

Cluster analysis reveals that searchers can be classified into three groups according to their magnitude of search: low (19% of the sample), high (11%) and "selective" (70% of the sample). Further, the composition of search, especially between the selective searchers appears to be quite different. Low searchers undertook little search from any source. These subjects were, however, most interested in cues from the Nonmarketer/Impersonal factor; the Nutritional Panel was the most frequently selected piece of information. High searchers made extensive use of most of the available sources, with six of the fourteen cues being selected by all subjects in this cluster. Cues from the Advocate factor were least selected by this group, while a government report on the product's artificial color received the least attention.

Examination of "selective searchers" reveals three distinct patterns of search, despite increasing magnitude across these groups. The low magnitude, selective searchers (23% of the sample) took most of the information sought from the Nonmarketer/Personal factor and secondly from the Nonmarketer/Impersonal group of cues. These people chose the least information from the Advocate group of cues, while the cue most frequently chosen was Family Comment, and the least was Sales Representative and Print-Ad. Using the language of adoption theory, this segment of searchers can be best labeled "Legitimizers" (Lionberger 1960), because of their reliance on nonmarketer information from impersonal and personal sources.

Sensitivity to Marketer/Impersonal and Nonmarketer/Impersonal information characterizes the search patterns of the intermediate selective searchers (27% of the sample). This group selected very little information from either the Advocate or Nonmarketer/Personal categories. Product Ingredients was most frequently selected while none of these searchers picked the Sales Representative cue. This group of people can best be labeled "Problem Solvers" because of their propensity to search for information mostly from Marketer/Impersonal sources, secondly from objective, testing sources of data and their avoidance of advocate and normative influence.

The final group of selective searchers (20% of the sample) is one which sought information mostly from Nonmarketer/ Personal sources, secondly from the Marketer/Impersonal group and a substantial amount from the Nonmarketer/Impersonal cues. This segment of searchers chose very little information from the Advocate group of cues, while every member of this group selected both the Family Comment and Friend's Comment information. Conversely, none of these decision makers selected the Sales Representative card. This segment of searchers can be best labeled "Normative/ Problem-Solvers," due to their "balanced" sensitivity to nonmarketer/personal, marketer/impersonal and nonmarketer/ impersonal types of information, coupled with their avoidance of advocate sources.

Correlates of Search

To contribute to improved understanding of these patterns of search, correlates of source use with three categories of variables were examined: demographics, task environment measures, and a behavioral measure (see Table 3). While the demographic variables are self-explanatory, the task environment and behavioral measures require brief discussion. The task environment group contains twelve items which measure selected aspects of the decision process to adopt. These include subjects' self-reported measures regarding their care and ease of food preparation general nutritional information use, bacon use and liking, brand switching, direct and indirect experience with TVP products and two measures of innovation proneness. The behavioral measure was based on the number of units purchased by the subjects.

The criteria for selecting these variables were (1) to include those commonly reported in other diffusion and information search studies which are reported to mediate magnitude and composition of search and (2) to also select those which were suited to the specifics of the product class (i.e., bacon) and the innovation (i.e., TVP). This exploratory, correlational analysis is used in identifying for future research, those variables which have the greatest apparent mediating effect, as well as determining whether these effects may differ between the f our information sources.

TABLE 3

CORRELATIONS OF SEARCH WITH INDEPENDENT VARIABLES

To document overall effects for this analysis, a one-way MANOVA was performed, using the five clusters as levels of an independent variable (search magnitude) with the 22 correlates as multiple dependent variables. Because of the exploratory nature of <his analysis, p-.10 will be considered trends, while p-.05 will be considered significant. A multivariate main effect of cluster was obtained (F=1.372, df=88,404; p=.0228) using the more conservative Pillia test. Follow-up univariate F-tests revealed that age and speaking to others who liked TVP had p<.01, while five other measures yielded p<.10. Following this overall test, zero-order correlations between the four indexes of search, total search, and the 22 variables were determined.

Demographic Correlates

Younger subjects selected more total information, as well as more from nonmarketer/impersonal sources, marketer/impersonal information, and nonmarketer/personal sources, while higher income was associated with greater levels of nonmarketer/personal search. Higher educational achievement lead to higher marketer/impersonal search, but a negative relationship with this type of information was found with higher spousal education. Higher home values correlated negatively with total search, as well as advocate and nonmarketer/impersonal source search The subject's social class had a positive relationship with marketer/impersonal search while spousal social class had a positive association with nonmarketer/impersonal source search. Finally, while married women searched for more information in total and from nonmarketer/impersonal sources but single women selected more marketer/impersonal information.

Task Environment and Behavior Correlates

Subjects who reported greater concern for cooking used more advocate sources, while those who claimed generally to use nutritional information sought more total information, more from nonmarketer/impersonal and advocate sources, but less from nonmarketer/personal sources. Women who stated a greater liking for bacon sought more in formation, comprised of more nonmarketer/impersonal source and marketer/impersonal information. The more different brands of bacon used (an indication of brand loyalty), subjects sought less marketer/impersonal information. Direct and conversational experience with TVP products had a negative relationship with total search, nonmarketer/impersonal, advocate, and nonmarketer/personal sources. Raving talked with others who liked TVP products lead to more marketer/impersonal search, but less nonmarketer/personal search. Those who claimed to be more generally aware of innovations sought more total information, more from nonmarketer/impersonal sources, more from the marketer/impersonal category, more advocate source information, but none with nonmarketer/personal sources. Women who purchased more of the TVP bacon used slightly more nonmarketer/impersonal source information. Surprisingly, bacon use, TVP familiarity and proneness to purchase food innovations before ones' friends had no effect on either pattern or magnitude of search.

Summary

These correlational results reveal the importance of both demographic and task environment variables in understanding search magnitude and composition. This can be seen in that the percentage of significant findings (39% demographics - 58: task environment) closely parallels the percentage of variables tested (41% demographics - 55% task environment). This conclusion argues for the necessity of diffusion theory researchers to carefully specify a broad range of relevant effects in their inquiries.

Another way to present the demographic and task environment findings presented above, is to examine the largest magnitude of effect for each of the four sources of information and total search. The Nonmarketer/Impersonal factor is best predicted with the extent to which individuals had talked to others about this class of innovations; this sociometric effect also had the largest relationship with two other issues: selecting advocate sources and total search. These findings underscore the importance of the role of sociometric effects in adoption of innovations. Finally, marketer/impersonal search is best explained by the extent of liking for the product category, while nonmarketer/personal search has the highest correlations with age.

DISCUSSION

To achieve the research goals of this study, especially insight into process, control of decision setting, and freedom from reliance on retrospection, sacrifices were made. Lack of mundane realism and possible threats to external validity are two which warrant discussion. Despite these sacrifices, the results generally mirror those found in studies using instrumentation other than the IDB technology. For example, despite a "cost free" information search environment, considerable variance in search magnitude occurred. Second, definitive search patterns emerged which generally parallel those reported in the adoption theory literature. Last, subjects' perceptions of information source developed into a typology closely approximating that which was expected. Consequently, these findings support the continued efficacy of IDB findings as argued by Lehmann and Moore (1980) and provide additional convergent validity to what is known of external search for innovations.

The findings of this study also underscore issues of continuing interest to adoption theory researchers, marketers, consumer educators and those concerned with public policy. For example, Robertson (1984) points to the need for increased use of market segmentation in improving understanding of the adoption process. The results of this analysis reveal clear distinctions across segments of searchers. Further study to more fully describe and determine the underlying reasons for these search segments would advance understanding of this aspect of adoption theory and confirm the value of examining the process using segmentation analysis.

These findings also confirm the wisdom of the marketer to use segmentation analysis. Marketers concerned with information sources under their control take great care to create formats, quantity and types of information to attract and persuade customers. The findings of this study affirm that different segments of decision makers display variable interest in the types of information controlled by marketers. Further, these results reveal that sociometric experience and product category liking are comparatively important mediators of search for marketer controlled types of information. This underscores the significance of task environment considerations to the marketer when developing communications strategies.

Consumer educators and those concerned with public policy need to continue their efforts to understand external search dynamics. Despite the low cost of search in this study. considerable magnitude variability was found.

Those who view search from a cost/benefit perspective, should be encouraged to shift their efforts toward the "benefit" side of this equation. The value of strategies developed to increase consumers willingness to search for more information becomes clear since this study confirms that many decision makers are not willing to search even in very low cost situations.

REFERENCES

Arndt, Johan (1967), "Word of Mouth Advertising and Informal Communication," in Risk Taking and Information Hand- ling in Consumer Behavior, ed. Donald F. Cox, Boston: Harvard University Press, 188-239.

Berning, Carol A. Kohn and Jacob Jacoby (1974), "Patterns of Information Acquisition in New Product Purchases," Journal of Consumer Research, 1 (Sept), 18-22.

Bettman, James R. (1970), "Information Processing Models of Consumer Behavior," Journal of Marketing Research, 7, 370-376.

Black, William (1983), "Discontinuance and Diffusion: Examination of the Post Adoption Decision Process," in Advances in Consumer Research, Vol. 10, et. Richard P. Bagozzi and Alice M. Tybout, Ann Arbor, MI: Association for Consumer Research. 356-361.

Dickerson, Mary Dee and James W. Gentry (1983), "Characteristics of Adopters and Non-Adopters of Home Computers," Journal of Consumer Research, 10 (Sept), 225-235.

Dixon, U. J. (ed) (1983), BMDP Statistical Software, Berkeley: University of California Press.

Duncan, Calvin P. and Richard W. Olshavsky (1982), "External Search: The Role of Consumer Beliefs," Journal of Marketing Research, 19 (Feb), 32-43.

Furse, David H., Girish N. Punj and David W. Stewart (1984), "A Typology of Individual Search Strategies Among Purchasers of New Automobiles," Journal of Consumer Re- search, 10 (Mar), 417-431.

Jacoby, Jacob (1984), "Perspectives on Information Overload," Journal of Consumer Research, 10 (Mar), 432-435

Katz, Elihu (1961), "The Social Itinerary of Technical Change: Two Studies on the Diffusion of Innovation," Human Organization, 20, 70-82.

Kiel, Geoffrey C. and Roger A. Layton (1981),"Dimensions of Consumer Information Seeking Behavior," Journal of Marketing Research, 18 (May), 233-239.

Lehmann, Donald R. and William L. Moore (1980), "Validity of Information Display Boards: An Assessment Using Longitudinal Data," Journal of Marketing Research, 17 (Nov), 450-459.

Lionberger, H. F. (1960), Adoption of New Ideas and Practices, Ames, Iowa: Iowa State University Press.

Newman, Joseph W. and Richard Staelin (1971), "Pre-Purchase Information Seeking for New Cars and Major Household Appliances," Journal of Marketing Research, 9 (Aug), 249-257.

Robertson, Thomas S. (1984), "Marketing's Potential Contribution to Consumer Behavior Research: The Case of Diffusion Theory," in Advances in Consumer Research, Vol. 11, ed. Thomas C. Kinnear, Ann Arbor, MI: Association for Consumer Research, 484-489.

Robertson, Thomas S. Joan Zielinski and Scott Ward (1984), Consumer Behavior, Glenview, Illinois: Scott, Foresman and Company.

Rogers, Everett M. (1976), "New Product Adoption and Diffusion," Journal of Consumer Research, 2 (Mar), 290-301.

Rogers, Everett M. (1983), The Diffusion of Innovations, New York: The Free Press.

Ryan, B. and N. C. Gross (1943), "The Diffusion of Hybrid Seed Corn in Two Iowa Communities," Rural Sociology, 8, 15-24.

Schaninger, Charles M. and Donald Sciglimpaglia (1980), "The Identification of Specific Information Acquisition Patterns in Information Board Tasks, and Their Relation to Demographics," Advances in Consumer Research, Vol. 7, ed. Jerry C. Olson, Ann Arbor, MI: Association for Consumer Research. 513-518.

Schaninger, Charles M. and Donald Sciglimpaglia (1981), "The Influence of Cognitive Personality Traits and Demographics on Consumer Information Acquisition," Journal of Consumer Research, 8 (Sept), 208-216.

Shoemaker, Robert W. and F. Robert Shoaf (1975), "Behavioral Changes in the Trial of New Products," Journal of Consumer Research. 2 (Sept), 104-109.

Wilton, Peter C. and Edgar A. Pessemier (1981), "Forecasting the Ultimate Acceptance of An Innovation: The Effects of Information," Journal of Consumer Research, 8 (Sept), 162-171.

----------------------------------------