|Home | About | Journals | Submit | Contact Us | Français|
Market or audience segmentation is widely used in social marketing efforts to help planners identify segments of a population to target for tailored program interventions. Market-based segments are typically defined by behaviors, attitudes, knowledge, opinions, or lifestyles. They are more helpful to health communication and marketing planning than epidemiologically-defined groups because market-based segments are similar in respect to how they behave or might react to marketing and communication efforts. However, market segmentation has rarely been used in alcohol research. As an illustration of its utility, we employed commercial data that describes the sociodemographic characteristics of high-risk drinkers as an audience segment; where they tend to live, lifestyles, interests, consumer behaviors, alcohol consumption behaviors, other health-related behaviors, and cultural values. Such information can be extremely valuable in targeting and planning public health campaigns, targeted mailings, prevention interventions and research efforts.
We describe the results of a segmentation analysis of those individuals who self-report consuming five or more drinks per drinking episode at least twice in the last 30-days. The study used the proprietary PRIZM™ audience segmentation database merged with Center for Disease Control and Prevention's (CDC) Behavioral Risk Factor Surveillance System (BRFSS) database. The top ten of the 66 PRIZM™ audience segments for this risky drinking pattern are described. For five of these segments we provide additional in-depth details about consumer behavior and the estimates of the market areas where these risky drinkers reside.
The top ten audience segments (PRIZM clusters) most likely to engage in high-risk drinking are described. The cluster with the highest concentration of binge drinking behavior is referred to as the “Cyber Millenials.” This cluster is characterized as “the nation's tech-savvy singles and couples living in fashionable neighborhoods on the urban fringe. Almost 65% of Cyber Millenials households are found in the Pacific and Middle Atlantic regions of the U.S. Additional consumer behaviors of the Cyber Millenials and other segments are also described.
Audience segmentation can assist in identifying and describing target audience segments, as well as identifying places where segments congregate on- or offline. This information can be helpful for recruiting subjects for alcohol prevention research, as well as planning health promotion campaigns. Through commercial data about high-risk drinkers as “consumers,” planners can develop interventions that have heightened salience in terms of opportunities, perceptions, and motivations, and have better media channel identification.
Audience segmentation has its origins in marketing research (Smith, 1956). Segmentation is a marketing technique which involves grouping audiences into segments or relatively homogeneous groupings, and treating each segment separately for product development, distribution, promotional, or communications purposes. The segments are usually derived from survey data, using techniques such as cluster analysis or latent class analysis to create relatively homogenous subgroups. Segmentation studies are based on the premise that audiences differ from one another, have different lifestyles, motivations and attitudes, and follow different patterns of behavior. In the product marketing world, these data are used to identify audience groups or segments of greatest opportunity.
Market or audience segmentation is used widely as a tool in social marketing efforts. Social marketing uses marketing principles and techniques to “influence a target audience to voluntarily accept, reject, modify or abandon a behavior for the benefit of individuals, groups, or society as a whole” (Kotler et al., 2002). Thus, it is frequently employed for public health, environmental and social causes. By grouping audiences into segments that share similar attitudes and behave in similar ways, it allows social marketers to develop communications strategies, interventions or services that closely match with the characteristics, requirements, and needs of individual segments (Slater and Flora, 1991). Despite the ubiquity of audience segmentation approaches in the business, communications, and public health world (Fridinger 2002, Dutta-Bergman 2004, Lyon-Daniel 2004, Pollard 2002, Robert 2004, and McDermott 2005) this typological method has rarely been employed in alcohol research. Yet, audience segmentation methodology provides insights into the behaviors, locations, social norms, lifestyle characteristics, cultural values and other traits of high-risk populations of drinkers with whom we would most like to intervene.
In this paper, we describe, as an illustrative example, the results of a segmentation analysis of high-risk drinkers using the proprietary PRIZM™ database merged with Center for Disease Controls' (CDC) Behavioral Risk Factor Surveillance System (BRFSS) database, as conducted by a marketing research company under contract to the National Institute on Alcohol Abuse and Alcoholism (NIAAA). PRIZM™ data is drawn from multiple data sources ranging from mailing lists, total census tract and postal data, household financial and insurance data, telecommunications, internet and satellite information, actual consumer purchasing (scanner) data, consumer marketing surveys and audience measurement systems, and existing proprietary marketing databases such as AC Nielsen, Gallup, IRI, JD Power, Mediamark, Nielsen Media Research, NFO, NPD, Polk Automotive, Scarborough and Simmons. These data are then typically aggregated into clusters that characterize various subgroups (i.e. segments) within the U.S.
Having reviewed a range of available audience segmentation databases with Kirby Marketing Solutions, we understood that each product offered valuable audience research information. However, considering our mission and the limited resources available for this purchase, we determined that Medstat's BRFSS/PRIZM merged dataset offered us the most valuable information to meet our most immediate needs at a reasonable cost. By integrating geographic, lifestyle and health information from a number of credible sources (BRFSS survey, census data, credit card purchases, DMV and other public records, etc.) this product provided information on lifestyle clusters, drinking habits, and geodemographic trends that could best inform current NIAAA activities, such as media campaigns, distribution of materials, constituent and congressional relations, and liaison partner selection. It also provided a context for any future NIAAA publications that may address binge drinking and/or young adults.
It is noteworthy that these audience data are not substitutes for more precise population rates of various attributes or conditions, such as NHIS or BRFSS studies of incidence or prevalence of high risk drinking. These data are however good approximations of more precise data (Pollard, 2002) and provide the extra benefit of describing the sociodemographic characteristics of high-risk drinkers as an audience; where they tend to live, their lifestyles, their interests, their consumer behavior, their alcohol consumption behavior, their other health-related behaviors, cultural values, and describe their relative size in the U.S. population. The richness of these data characterizing high-risk drinkers clearly lends itself to the generation of hypotheses concerning putative public health interventions that might target specific populations or locations in order to attenuate this pattern of risky alcohol consumption, and reduce its associated public health burden. In addition, the results are informative from the perspective of research subject recruitment by describing those settings where problem drinkers are most likely to be found, and informing on how to best reach this audience.
These analyses was conducted using the proprietary PRIZM™ (Claritas, Inc., San Diego, California) syndicated audience segmentation system merged with the CDC's BRFSS database purchased from Thomson Medstat, Inc. (Ann Arbor, MI). The analyses were conducted by Kirby Marketing Solutions (SDK). PRIZM™ segmentation divides all U.S. households into 66 clusters; starting with stratification by three affluence levels (high, middle and low), then layering on population density based upon four distinct categories (urban, suburbs, 2nd cities, and town & country), resulting in 12 social groups. The geographic nature of the data enables users to identify clusters at the national, state, country, city, zip code and block group (~300 households). Thus, users can identify where their target audiences live, where they work, shop, received medical care, as well as characterizing the communities in which they live.
Consumer information is merged with the demographic dataset. PRIZM™ integrates proprietary data from the Simmons Market Research Bureau (Ft. Lauderdale, Fla). Simmons measures over 30,000 American consumers each year using statistically projectable samples. Simmons' samples include the largest representation of all Americans, including, Hispanic, African American, Asian and other consumers and are used by over 500 clients worldwide to provide the most complete picture available on America's consumers. PRIZM™ provides a core database of all U.S. census demographic data, public records (such as Department of Motor Vehicle data), private sector purchasing databases (such as credit card and store scanner data), and Simmons consumer data. PRIZM™ then categorizes all U.S. households into 66 clusters based on based upon affluence, geography, population density, and consumer and lifestyle data. Clusters account for 93% of the variance in the data, and are identified with folksy nick-names. Detailed descriptions of the characteristics of these 66 clusters may be found on the Claritas, Inc. website: http://www.claritas.com/MyBestSegments/Default.jsp. The authors have changed the proprietary names of the Claritas clusters in accordance with licensing requirements.
The Behavioral Risk Factor Surveillance System (BRFSS) is a state-based system of health telephone surveys that collects information on health risk behaviors, preventive health practices, and health care access primarily related to chronic disease and injury. It was started by the Center for Disease Control in 1984. Currently, data are collected monthly in all 50 states, the District of Columbia, Puerto Rico, the U.S. Virgin Islands, and Guam. More than 350,000 adults are interviewed each year for the BRFSS.
For the analyses reported herein, the 2004 BRFSS national data (based on the 2004 U.S. census) on high-risk drinking was merged with the PRIZM™ cluster database. PRIZM™ and BRFSS data were merged using a proprietary algorithm based on the demographics of each PRIZM™ cluster, which is then “reverse engineered” from BRFSS to a cluster assignment using detailed respondent demographics. The model was tested multiple times with multiple similar items between datasets to ascertain the best matching algorithm. The algorithm itself is proprietary and unavailable for dissemination but is a process used successfully and widely by the vendor to merge private sector datasets that do not provide a direct street address for importing into PRIZM™.
We utilized the BRFSS question “considering all types of alcoholic beverages, how many times during the past 30 days did you have 5 or more drinks on one occasion” as the index measure of high-risk drinking. This measure was chosen because it reflects the level of ethanol intake at which impairment in psychomotor and cognitive functioning is significantly increased (Dawson et al., 1996) and numerous adverse social, medical, occupational, and legal outcomes of drinking increase (e.g. Greenfield, 2001; Rehm et al., 2005; Wechsler and Nelson, 2006), as the frequency of consuming 5+ drinks rises. In addition, it is a standard BRFSS consumption measure, has multiple years of data, has been reliability and validity tested, and was available to be merged into the PRIZM™ dataset. The Claritas PRIZM™ cluster provided a framework into which all clusters were placed. This framework consists of Lifestage, age and presence of children, as well as Social Groups, a measure of urbanization. The first step in cluster assignment is the designation of the Social Group and Lifestage for each survey respondent contained in the BRFSS dataset.
The additional characteristics used to further differentiate clusters were:
An analysis of the national averages of the cluster demographics was used to determine the cut points for the variables used to determine the final cluster assignments. Each record was then assigned a cluster based on its own particular combination of Lifestage, Social Group, income, education, race and ethnicity.
Our initial evaluation demonstrated that 14.9% of the BRFSS sample responded affirmatively to having a heavy drinking episode once in the last thirty days. In order to focus on a higher-risk drinking grouping, we used a threshold of having at least two such 5+ drinking episodes in the last 30 days, as endorsed by 6.1% of the BRFSS sample. This better serves our purpose of identifying characteristics of individuals at greatest risk for the consequences of heavy drinking.
We report herein on the audience clusters that are 50% or more likely than the US average to engage in high-risk drinking at least twice in the last 30 days. This metric is based on the calculation of a proprietary index of concentration (IOC) measure. Such concentration measures are widely used tool in marketing research, although no universally accepted market concentration measure exists (Xu, 2005). In this case, the IOC is a cluster's relative propensity to engage in a behavior, compared to all other clusters. The procedure is as follows. The number of people within a cluster who self-report a behavior is divided by the total number of people within that cluster. This provides a proportion of those behaving versus the entire cluster. For example, Cyber Millenials consisted of 2,538,595 people in 2004 and 314,219 (12.4%) Cyber Millenials reported binge drinking at least twice in the last 30 days. The IOC then compares binge drinking for a cluster to the US average. In the U.S. 6.1% of all respondents reported binge drinking in 2004 and 12.4% of Cyber Millenials reported binge drinking, thus the IOC is 203 (12.4%/6.1%). Cyber Millenials members are therefore 2.03 times more likely to be binge drinkers than the US average. In this sense, the IOC acts much like a relative risk measure in public health.
NIAAA staff then selected a subset of five of the ten clusters for more in-depth analysis. We selected this subset primarily due to cost constraints of purchasing full PRIZM™ profiles for all ten clusters that reported binge drinking at 150% or more of the US average. We began the process by selecting clusters that had a youth focus because 1) a youth focus is consistent with NIAAA programming and 2) youth represent an important time period in which to intervene to minimize the most damage done by excessive alcohol intake. In addition, 3) these 5 clusters represent 20% of all binge drinkers (a sizeable group to reach), and these five clusters have one and half to two times the average propensity to have members who are binge drinkers. We also reasoned that NIAAA could more cost-effectively plan for the purchase of mass media, social media, or community-based outreach if there were common communication channels among a subset of clusters. This is not to suggest that the messages might be the same for all of the clusters in the subset, only that common channels might offer some economies of scale.
Table 1 displays the top ten market segments wherein individuals have endorsed having 5 or more drinks on one occasion at least twice in the last 30 days ranked according to IOC scores. These top ten high-risk drinking clusters, in order of endorsement rates and IOC scores, are described as follows:
As discussed previously, NIAAA has an interest in youth-oriented outreach, has research to suggest that youth benefit the most from early outreach, and youth behaviors are less entrenched than adults, so more amenable to change. NIAAA was also keenly aware that funding for tailored outreach for each of the five youth-oriented PRIZM™ clusters was unlikely. It was more likely that funding would be available to support one or two programs. Therefore an analysis of common factors across the five youth clusters was conducted to identify potential opportunities for aggregating PRIZM™ clusters into larger but similar target groups. Specifically, we selected the “Cyber Millenials”, “Avant-Garde Mix”, “City Producers”, “Metro Newbies” and “Mobile Ladder Climbers” for further study. The predominant demographic and geographic characteristics of these five market segments are displayed in Table 2. See Figure 1 for a diagram of the decision making process.
The “Avant-Garde Mix” and “Cyber Millenials” share similar education, occupation, geography, ethnic mix, life stage, social group, income factors, and a myriad of consumer and lifestyle behaviors. The remaining three clusters, “City Producers”, “Metro Newbies”, and “Mobile Ladder Climbers” share similar education, life stage, occupation, geography, incomes, and consumer and lifestyle behaviors. As can be seen, two of the clusters have members predominantly under age 35 and three have members predominantly up to age 44. All five clusters represent urban environments in varying degrees, with two also encompassing towns and suburban fringe areas. Four of the clusters are geographically located in Pacific, Mid-Atlantic, East-Central, and South-Pacific coastal regions, while one cluster has no specific geographic attributes. Although each of the five selected clusters has a predominance of White members, the clusters are also overrepresented by ethnically diverse populations.
In the media marketing world, Designated Market Areas (DMA) are large geographic coverage areas for mass media markets. Specifically, they are regions where the population can receive the same or similar television, radio station offerings and may include newspapers and internet content. As can be seen in Table 3, there are eleven DMAs that cover 47% of all of the five high-risk drinking clusters (3,776,029 individuals). These 11 DMAs can provide an efficient and focused approach to reaching target populations. These large urban DMAs provide reach to high-risk drinking clusters in the following proportions:
These geographic regions could be suitable venues for transit ads, DMA-specific magazine, radio, and cable advertising, community prevention campaigns, research recruitment, or even identifying high potential areas for health agency collaboration.
PRIZM™ data also provides mass media usage information for cable and network TV usage, specific TV show viewership, radio format and time of day listening, and newspaper and magazine readership. Table 4 shows the 2004 activities with the high Index of Concentration (IOC > 150) and double digit proportions of the cluster usage for each media source for the high-risk drinking segments. The common activities row in the Table indicates media activities that are common to two or more clusters, representing an opportunity to reach multiple clusters with the same mass media elements.
The alcoholic beverage preferences for all five high-risk drinking clusters were also examined. We selected beverage preferences that were high in concentration (>150) and high in participation (>10%). The beverage choices common to two or more high-risk drinking clusters are displayed in Table 5 along with cluster specific preferences. Knowing the actual beverage preferences of high-risk drinking clusters can provide data that can be used for targeted communication, community outreach, and even possibly industry partnerships.
It is noteworthy that “Metro Newbies” and “Mobile Ladder Climbers” have elevated rates of tobacco smoking beyond that of the U.S. population, while “Cyber Millenials” have lower smoking rates than the U.S. population. “Mobile Ladder Climbers” are also more likely than the U.S. average to have tried to quit smoking using multiple methods. These data also suggest that for some high-risk drinking segments, prevention communications could address drinking and smoking behavior simultaneously.
Members of these high-risk drinking clusters use the Internet at higher rate than the US average. While the Cyber Millenials are the most tech-savvy cluster, the other clusters also use the Internet for connecting with people, searching for information, reading magazines, and shopping at high rates. The common behaviors are listed in the last row of Table 6. These data suggest that prevention outreach strategies or subject recruitment efforts should include an Internet component.
The high-risk drinking clusters do not appear to participate in sports in large numbers, but some clusters do watch/listen to sports coverage (basketball, MLB, and NASCAR). As can be seen in Table 7, one common denominator among three of the clusters is having a liberal outlook at rates much higher than the US average. Buying music and books is also another common leisure behavior. Finally two of these clusters share an interest in Yoga and exercising at a health club. These leisure activities suggest that outreach to favorite musical artists and self-help book authors may prove effective as spokespeople for a high-risk drinking reduction campaign or other prevention activities.
In this report, we have demonstrated how proprietary audience segmentation data can be linked to publicly-accessible epidemiological survey data to inform about the demographics, location, and consumer behaviors of individuals who engage in high-risk drinking practices or other risky behaviors. Such information could then be used as the basis for a health communications campaign, targeted distribution of informational material, public health outreach activities, community-based prevention interventions, prevention research venue selection, and even research subject recruitment strategies. Using a definition of high-risk drinking behavior (e.g. more than 5 drinks per occasion at least twice in the last 30 days), we have described the sociodemographic characteristics of these high-risk drinkers as an audience segment; where they tend to live, their lifestyles, their interests, their media preferences, their consumer behavior, their leisure activities and their other health-related behaviors. The results have allowed us to hypothesize about components of a public health intervention that might target these specific populations (i.e. audience segments) in order to reach these populations and intervene in a manner that might attenuate their pattern of risky alcohol consumption.
The utility of this approach lies in its efficiency from a cost and effort perspective. Comprehensive, community-based approaches have emerged as the most viable way of reducing the risk of alcohol and other drug use (Aguirre-Molina and Gorman, 1996; Dryfoos, 1993). Rather than engaging in a broad general prevention intervention, audience segmentation permits focusing on those subpopulations within communities at greatest risk, while it simultaneously affords useful consumer market data to tactically penetrate those high-risk sectors. While unitary media interventions in the past appeared to have had little impact on the use of alcohol, drugs and cigarettes (Flynn et al., 2006), successful media campaigns have been demonstrated to have greatest utility when integrated into broad-brush community-based prevention interventions (Aguirre-Molina and Gorman, 1996; Bryant et al., 2007) (Johnson et al., 1990) (Brannon et al., 1989) (Flay, 1986).
The incorporation of audience market information into prevention and health promotion efforts is a component of “social marketing”. Social marketing, as defined by Kotler and Andreasen (1991) (Kotler and Andreasen, 1991) “seeks to influence social behaviors not to benefit the marketer, but to benefit the target audience and the general society.” This concept has been extended to the marketing of positive health behaviors as public health interventions (Lefebvre and Flora, 1988). Important concepts in social marketing approaches include (Hastings and Haywood, 1991): consumer orientation; market segmentation and the identification of target markets; objective setting, the formulation, implementation monitoring and evaluation of the “marketing mix” (i.e. the health product, its cost, its promotion and distribution). In this report, we have only identified the target market (i.e. high-risk drinkers) employing the “consumer” perspective. The market objective of a reduction in high-risk drinking behavior has only been alluded to, and we have hypothesized a formulation based upon market information. The implementation of such a health promotion initiative is beyond the scope of this manuscript.
There are several limitations of this market segmentation approach. First, it emphasizes the use of established and proprietary market segments. Due to the proprietary nature of the databases employed, their formulation is not open to public scrutiny or scientific validation in the concurrent or predictive domains. Thus, we cannot comment on the validity of the clusters, the matching procedures or other technical points that represent the “black box” of purchased proprietary services. Second, the market segment is the unit of analysis, not the high-risk drinker within that segment. Thus, our capacity to identify characteristics of the high-risk drinker is limited to descriptions of the segment or cluster in which he or she is subsumed. Likewise, changes produced by an intervention will only be reflected in group drinking behavior, not individual differences. The third limitation is the cross-sectional nature of this approach. We employed a single index year (2004) of BRFSS data for our analyses. Thus, demographic and behavior changes since that year may reduce the accuracy of our market segmentation approach. Likewise, the characterization of audience segments by the PRIZM™ database may be markedly different in 2008. Despite these challenges and issues, we believe this process hold promise for the future because mass marketing has proven to be substantially less effective than targeted marketing, such as this example.
We hope this report has demonstrated how audience segmentation can assist in targeting venues and populations for alcohol prevention research, as well as health promotion campaigns. Through insights about the behavior of high-risk drinker as “consumer”, one can craft interventions that have heightened salience in terms of opportunities, perceptions, and motivations.
This research was supported by internal NIAAA funds.