Cognitive age

In the past, age was frequently used as a segmentation base⁽ Reference Lumbers, Raats, Shepherd and Raats ^{37 ,} Reference Bone ^{38 )}. However, since the 1990 s, marketers have started to acknowledge that people do not ‘act their age’ and that people within the same age group are rather dissimilar⁽ Reference Moschis ^{6 )}. Chronological age is an empty variable, in the sense that it does not cause behaviour⁽ Reference Settersten and Mayer ^{39 )}, but merely gives a suggestion of one's degree of ageing. Age, in turn, is not perfectly in line with ageing⁽ Reference Sudbury and Simcock ^{36 )}. Ageing occurs at different rates in different people and elderly often feel much younger than their chronological age⁽ Reference Mathur and Moschis ^{40 )}. As a result, chronological age has made way for cognitive age; the age one feels like, looks like, acts like and the age one's interests fit in⁽ Reference Bone ^{38 ,} Reference Barak and Schiffman ^{41 )}.

Cognitive age results from biological, psychological and social ageing, which affect how people perceive themselves⁽ Reference Mathur and Moschis ^{40 )}. After retirement, for example, people may start feeling old because they want to maintain consistency between their behaviour and their identity⁽ Reference Mathur and Moschis ^{40 )}. Nonetheless, the majority of elderly do not feel old⁽ Reference Barak and Rahtz ^{42 )}, which may be partly explained by the perceived undesirability of old age that exists in many societies⁽ Reference Tepper ^{16 ,} Reference Barak and Rahtz ^{42 )}. Also, the turning point from feeling young to feeling old may require the experience of a catalytic event, rather than ageing in general⁽ Reference Van Auken and Barry ^{43 )}, and the timing of this event probably differs considerably among individuals because people respond differently to life events⁽ Reference Brownie ^{9 ,} Reference Mathur and Moschis ^{40 ,} Reference Delaney and McCarthy ^{44 )}. However, as long as one lives in a stable environment, cognitive age may remain constant or change slowly⁽ Reference Mathur and Moschis ^{40 )}. In line with this, the test–retest reliability of the cognitive age scale is strong; 0·88⁽ Reference Bone ^{38 ,} Reference Barak and Schiffman ^{41 )}.

Cognitive age mediates the effect of various demographics on behaviour⁽ Reference Mathur and Moschis ^{40 )} and it may therefore provide a rich perspective of consumers, resulting in uniquely responsive segments. Cognitively old elderly, for example, are more price sensitive than cognitively young elderly⁽ Reference Barak and Gould ^{45 )}. Given the overall higher prices of functional food^{( 46 )}, cognitively old elderly may thus be less interested in purchasing these types of food than cognitively young elderly. However, to our knowledge, there is no literature available on the relationship between cognitive age and (functional) food choice.

Cognitive age is a better predictor of behaviour than chronological age⁽ Reference Van Auken and Barry ^{43 )}. Nevertheless, as a person-level variable, cognitive age is limited in predicting food choice. In addition, segments based on cognitive age are not as easily identified as segments based on chronological age. Whereas chronological age can be determined objectively and directly, using a single question, cognitive age is inferred from responses on a multi-item questionnaire, albeit a small and reliable one (split-half reliability: 0·85)⁽ Reference Barak and Schiffman ^{41 )}. Moreover, there is no consensus on how to categorise subjects into cognitive age groups. One could, for example, use a cut-off score to categorise people into cognitively young v. old, or one could distinguish between people who feel younger or older than their actual age. Although this lack of consensus reduces identifiability of segments based on cognitive age, it does provide marketers with some influence on the composition and substantiality of segments; there is no fixed number of segments (in contrast with, for example, sex).

Irrespective of the operationalisation of its groups, cognitive age provides limited instructions for marketing efforts. Questions may be raised such as: do elderly who feel like 60 want different functional foods from those who feel like 90? Actionability for product design is thus limited and, therefore, researchers usually recommend combining cognitive age with other variables⁽ Reference Sudbury and Simcock ^{36 ,} Reference Van Auken and Barry ^{43 )}. Segments do provide some instructions for communication efforts and are moderately actionable for communication. For example, research on age-related stigmatisation has shown that cognitively old elderly accept age-targeted products and services such as senior discounts, whilst cognitively young elderly do not⁽ Reference Tepper ^{16 ,} Reference Moschis and Mathur ^{47 )}. Furthermore, media profiles based on chronological age may provide some information on how marketers can access segments based on cognitive age, since they are correlated (r>0·50)⁽ Reference Mathur and Moschis ^{40 ,} Reference Barak and Rahtz ^{42 )}. Media profiles based on other demographics may not be applicable because they are only weakly related to cognitive age⁽ Reference Mathur and Moschis ^{40 ,} Reference Van Auken and Barry ^{43 )}.

Life course

Whereas cognitive age perspectives assume that life events affect one's self-perception, life course perspectives argue that elderly who experienced similar life events will have similar needs, wants and/or behaviours⁽ Reference Moschis ^{6 ,} Reference Moschis ^{21 )}. Life events like retirement, remarriage, becoming a grandparent, relocation, health problems and widowhood⁽ Reference Sobal and Bisogni ^{48 )} have all been studied in the context of food consumption, but their effects tend to vary from one person and situation to another⁽ Reference Delaney and McCarthy ^{44 ,} Reference Mathur, Moschis and Lee ^{49 )}. Retirement, for example, may provide people with more time to prepare and enjoy meals, but may also increase the tendency to snack out of boredom or inactivity⁽ Reference Delaney and McCarthy ^{44 )}. This variability can be explained by the idea that life events have various direct and indirect effects⁽ Reference Mathur, Moschis and Lee ^{49 )} that interact with each other in a dynamic way⁽ Reference Delaney and McCarthy ^{44 )}.

Because life events cannot be viewed in isolation⁽ Reference Mathur, Moschis and Lee ^{49 )}, most researchers use combinations of life events. One example is the gerontographic life-stage model, which is specifically aimed at segmenting the elderly population. The model proposes four substantial segments: healthy indulgers (13 %), healthy hermits (38 %), ailing outgoers (34 %) and frail recluses (15 %)⁽ Reference Moschis ^{21 )}, which consist of elderly who are at different but not necessarily sequential stages in their lives⁽ Reference Lumbers, Raats, Shepherd and Raats ^{37 )}. In part because segments are based on a wide collection of measures, the life-stage model predicts behaviour better than chronological age, cognitive age⁽ Reference Moschis ^{21 ,} Reference Moschis, Curasi and Bellenger ^{50 )} or demographics⁽ Reference Moschis, Curasi and Bellenger ^{51 )} alone and segments are differentially responsive. Ailing outgoers, for example, are interested in learning new things, whilst frail recluses want to feel secure⁽ Reference Moschis, Curasi and Bellenger ^{50 )}. However, the extent to which different marketing strategies appeal to the segments varies across products⁽ Reference Moschis ^{21 )} and segments are, therefore, limited in actionability regarding both product design and communication. In addition, it has been suggested that only two segments are viable for the food market: healthy indulgers and ailing outgoers⁽ Reference Moschis, Lee and Mathur ^{52 )}.

In contrast, stability of segments based on life course perspectives is probably good, because life events are closely linked with demographics (for example, retirement with income and occupation), which are generally stable. In line with this, the gerontographic life stages are relatively stable, with segments varying in size by no more than 2 % across studies⁽ Reference Moschis ^{21 )}. Moreover, some of the demographic information inherent in life events can be linked to media profiles (for example, socio-economic status), which provides marketers with information on the accessibility of segments. Although single life events may be easily identifiable using objective and direct demographic questions, questionnaires that assess a collection of life events can become quite lengthy. In addition, there is considerable variance among studies in the choice of life events, and concrete information on the variables used to measure the gerontographic life stages is lacking⁽ Reference Weijters and Geuens ^{53 )}, because the model is proprietary⁽ Reference Faranda and Schmidt ^{17 )}.

Time perspective

Socio-emotional selectivity theory is another approach that looks at the effect of life events on behaviour⁽ Reference Cate and John ^{54 )}. Rather than focusing on self-perception (like cognitive age) or adaptation (life course perspectives), it assumes that differences in behaviour are based on differences in time perspective⁽ Reference Carstensen, Fung and Charles ^{55 )}. The theory classifies time perspective into limited time and open-ended time, which can be measured using a reliable, ten-item questionnaire (internal consistency: 0·92)⁽ Reference Lang and Carstensen ^{56 )}. People who perceive time as limited (regularly found in ill or old people) have a tendency to pursue short-term, emotional and present-oriented goals whereas people who perceive time as open-ended (more often found in young people) have a tendency to pursue more long-term, rational and future-oriented goals⁽ Reference Carstensen, Fung and Charles ^{55 ,} Reference Ares, Giménez and Gámbaro ^{57 )}.

On a consumer level, hedonic products (for example, tasty food) are most appealing when time is limited, whereas utilitarian products (for example, healthy food) are most appealing when time is open-ended⁽ Reference Wei, Donthu and Bernhardt ^{58 )}. Functional foods usually provide benefits in the long run and may therefore, on the one hand, appeal most to consumers with an open-ended time perspective. On the other hand, the convenience or ‘quick fix’ aspect of functional foods⁽ Reference Cornish ^{59 )} may appeal to those who perceive time as limited. In line with the latter, elderly living in nursing homes (i.e. who probably perceive time as limited) are more interested in functional foods than those still living independently⁽ Reference Van Der Zanden, Van Kleef and De Wijk ^{60 )}. However, to our knowledge, no published studies have looked at the effect of time perspective on functional food acceptance.

Although time perspective is correlated with age, such that older adults more often perceive time as limited, it is not bound to age⁽ Reference Carstensen, Fung and Charles ^{55 ,} Reference Wei, Donthu and Bernhardt ^{58 )}. In fact, larger differences in time perspective exist among elderly than among young adults⁽ Reference Fung, Lai and Ng ^{61 )} and the correlation between age and time perspective is not always significant within the elderly population⁽ Reference Bal, Jansen and Van Der Velde ^{62 ,} Reference Fredrickson and Carstensen ^{63 )}. In fact, time perspective can even be changed temporarily using time constraints or statements⁽ Reference Wei, Donthu and Bernhardt ^{58 ,} Reference Kim, Goldstein and Hasher ^{64 )}. Segments based on time perspective are therefore more or less stable depending on the context of food choice. In line with this, test–retest reliability of the future time perspective scale is moderate (0·50 to 0·72)⁽ Reference Zacher and De Lange ^{65 )}. In addition, as elderly more often perceive time as limited, a segment of elderly who perceive time as open-ended may not be substantial enough to be profitable. In addition, time perspective is not strongly related to demographics other than chronological age⁽ Reference Cate and John ^{54 )}, and this association is weak within the elderly population⁽ Reference Bal, Jansen and Van Der Velde ^{62 ,} Reference Fredrickson and Carstensen ^{63 )}. As a result, information on accessibility of segments based on time perspective will be limited.

Time perspective is, however, strongly related to one's motivations and goals⁽ Reference Carstensen, Fung and Charles ^{55 )} and may therefore provide responsive segments. Emotionally appealing advertisements are, for example, more attractive to consumers who perceive time as limited than to consumers who perceive time as open-ended⁽ Reference Williams and Drolet ^{66 )}. In addition, time perspective is inferred from responses on a validated, multi-item questionnaire and is therefore moderately easy identifiable. Moreover, segments based on time perspective are actionable for both communication and design, because they provide information about the products and messages that are appealing to consumers (for example, hedonic v. utilitarian). A recent conceptualisation of time perspective may provide even more actionability. Cate & John⁽ Reference Cate and John ^{54 )} found that time perspective could best be operationalised as two independent dimensions: perceived limitations and opportunities. In this conceptualisation, a person who sees time as limited does not necessarily see time as less full of opportunity and may be interested in, for example, products that promote health and are convenient to use at the same time.

Demographics

In contrast with time perspective, demographics are widely used as segmentation bases. They provide segments that are considerably stable (for example, sex), substantial (for example, education), easily identified using objective measures (for example, age)⁽ Reference Van Raaij and Verhallen ^{27 )} and intuitively easy to understand⁽ Reference Honkanen, Olsen and Myrland ^{67 )}. In addition, demographics are often readily available⁽ Reference Wedel and Kamakura ^{22 )} and have been used as a basis for media profiles that provide marketers with information on the accessibility of consumers.

Functional food acceptance has been related to various demographics, such as sex, education, income and age⁽ Reference Childs and Poryzees ^{68 ,} Reference Verbeke ^{69 )}. Specifically, women and older adults are overall found to be more accepting of functional foods than men and younger adults⁽ Reference Poulsen ^{18 ,} Reference Bogue and Ryan ^{70 )}, and consumers with lower education have more concerns about functional foods than those with higher education⁽ Reference Niva and Mäkelä ^{71 )}, which may also be explained by their lower income. These findings suggest that highly educated, wealthy, elderly women may be most interested in functional foods.

Nevertheless, differences in demographics generally do not account for much variation in actual food choice⁽ Reference Rozin, Shepherd and Raats ^{72 ,} Reference Locher, Ritchie and Roth ^{73 )}. For example, the core foods eaten by elderly vary little with income⁽ Reference Fanelli and Stevenhagen ^{74 ,} Reference Briley ^{75 )} and the perceived health benefits of functional foods outweigh the effects of age and sex on functional food acceptance⁽ Reference Verbeke ^{69 )}. Furthermore, segments based on demographics are limited in actionability regarding both product design and communication⁽ Reference Steenkamp and Ter Hofstede ^{35 )} as they provide little instructions for marketing efforts. For example, how can marketers position functional food in a segment of well-educated, elderly women? In addition, segments are probably not uniquely responsive to marketing efforts because consumers within these segments are often considerably heterogeneous⁽ Reference Haley ^{26 )}. Indeed, relationships found between demographics and food choice differ between functional foods⁽ Reference Ares and Gámbaro ^{76 )}, across countries⁽ Reference Payette and Shatenstein ^{77 )}, over time⁽ Reference Verbeke ^{78 )} and across studies in general.

Summary

Person-level segmentation bases are generally stable, responsive and substantial. They are, however, limited in their actionability because they are not strongly related to actual food choice. Therefore, these bases provide limited instructions for marketing efforts in the functional food market. Furthermore, demographics and single life events are unique in the sense that they are measured directly with objective, one-item measures (for example, birth date, sex). All other segmentation bases that we discuss are inferred from multi-item questionnaires.

General food beliefs

The experiences with food that consumers have over their lifetime affect their personal beliefs about food, such as what types of food they should or should not eat, the role of food in their lives and how they categorise or evaluate food products⁽ Reference Rozin, Shepherd and Raats ^{72 ,} Reference Locher, Ritchie and Roth ^{73 ,} Reference Herne ^{80 ,} Reference Sobal, Bisogni, Devine, Shepherd and Raats ^{81 )}. These beliefs are general, in that they are not about specific food products (for example, oranges) but about broader food categories (for example, fruit) or about food in general (for example, the effects of food on health). General food beliefs differ strongly among consumers and, as a consequence, the same type of food (for example, functional food) may be perceived as healthy and convenient by some, but as unhealthy and quick-fix solutions by others⁽ Reference Sobal, Bisogni, Devine, Shepherd and Raats ^{81 )}.

Elderly tend to believe that their diet is already healthy⁽ Reference De Almeida, Graça and Afonso ^{82 ,} Reference Kremer, Hulst and Boesveldt ^{83 )} and that improving their diet is too late⁽ Reference Delaney and McCarthy ^{44 ,} Reference Herne ^{80 )}. Single beliefs are probably not actionable, however, because they do not directly translate into specific behaviours. For example, consumers may overall be sceptical about functional foods, but nevertheless consume specific functional foods that meet their needs or wants⁽ Reference Van Der Zanden, Van Kleef and De Wijk ^{60 )}. A combination of multiple beliefs thus provides a broader perspective of consumer thought processes and food choice. In one of the few segmentation studies of the elderly food market, Morgan⁽ Reference Morgan ^{23 )} segmented elderly using a questionnaire with food-related statements (for example, ‘eating at restaurants is too expensive’) and found three responsive segments: nutrition concerned, fast and healthy, and traditional couponers. Segments like these are actionable for both product design and communication, because they provide concrete information about the thought processes of consumers. For nutrition-concerned elderly, for example, marketers can emphasise the health aspects of foods and for fast and healthy elderly the convenience aspects.

One well-known questionnaire based on food beliefs is the Food-Related Lifestyle Instrument. It measures sixty-nine food beliefs from the process of food shopping to food consumption and links them to five food lifestyle elements⁽ Reference Grunert, Brunsø and Bisp ^{84 )}. However, most researchers construct their own questionnaire, with beliefs that are of central interest for specific research questions. Similar to measuring life events, there is no consensus on which food beliefs should be measured. Nonetheless, the choice of beliefs is a critical step⁽ Reference Hair, Black, Babin, Hair, Black, Babin and Anderson ^{25 )} that strongly affects actionability. A segmentation based on beliefs about food in general will provide less information on how a functional food should be designed (actionability for design) but more information on how to approach consumers (actionability for communication) than a segmentation based on beliefs about the more specific category of functional foods.

Substantiality of segments is probably not a problem when using food beliefs, because marketers are not bound to a specific number or type of segments and can thus combine segments that are too small. Furthermore, general food beliefs are based on a lifetime of experiences and can remain consistent over long periods of time⁽ Reference Edstrom and Devine ^{85 )}. In contrast, the more specific beliefs about functional foods may not yet be strongly established in consumers' minds. In earlier studies on functional foods, beliefs used to fluctuate between as well as within experiments⁽ Reference Urala and Lähteenmäki ^{86 ,} Reference Landström, Hursti and Magnusson ^{87 )} and although functional foods are more common nowadays, consumers are still ambivalent about them⁽ Reference Van Der Zanden, Van Kleef and De Wijk ^{60 ,} Reference Coleman, Miah and Morris ^{88 )}. Segments based on food beliefs may thus be more or less stable depending on the beliefs that are measured. The availability of media profiles and, therefore, accessibility of the segments also depends on the choice of beliefs.

Food choice motives

Whereas food beliefs consist of what consumers think is true, food choice motives consist of what consumers think is important and which factors they consider in making food choices.

The widely used Food Choice Questionnaire⁽ Reference Steptoe, Pollard and Wardle ^{89 )} assesses the importance of each of nine food choice motives: health, mood, convenience, sensory appeal, natural content, price, weight control, familiarity and ethical concern (internal consistency of the scales>0·72). A later adjustment of this scale divides the ethical motive into three motives: religion, political value and ecological welfare⁽ Reference Lindeman and Väänänen ^{90 )}. Segments based on food choice motives are actionable for both product design and communication, because the motives can be directly translated into product characteristics and messages⁽ Reference Kornelis, Van Herpen and Van Der Lans ^{91 )}. Segments that are motivated by weight control, for example, may be interested in protein-enriched products that ‘increase feelings of satiety’ whereas segments that are motivated by natural content may prefer products that ‘contain natural sources of protein’.

Usually consumers hold multiple motives⁽ Reference Kornelis, Van Herpen and Van Der Lans ^{91 )} that may be more or less important depending on the context of food choice⁽ Reference Onwezen, Reinders and Van Der Lans ^{92 )}. When hospitalised, for example, people often temporarily care less about the sensory appeal of food and focus more on their health⁽ Reference Herne ^{80 )}. In the case of conflicting motives, a trade off takes place, wherein one can, for example, focus on a single motive (for example, always choose the healthiest option). Alternatively, one can routinise ways to trade off motives for recurring situations (for example, only choose tasty foods during the weekend)⁽ Reference Sobal, Bisogni, Devine, Shepherd and Raats ^{81 )}. In addition, consumers may choose to interpret food choice motives in such a way that they are consistent with their behaviour. When consuming ice cream, for example, consumers may operationalise their health motives in terms of promoting wellbeing (i.e. enjoyment) even though the product may not benefit their physical health⁽ Reference Chrysochou, Askegaard and Grunert ^{93 ,} Reference Van Dam and Van Trijp ^{94 )}. Furthermore, what consumers think is important does not always align with what they actually do⁽ Reference Urala ^{95 )}. Food choice motives are thus probably not strongly predictive of actual food choice.

Trade-off tendencies of food choice motives are relatively stable within situations⁽ Reference Sobal, Bisogni, Devine, Shepherd and Raats ^{81 )} and, usually, food choice questionnaires measure the importance of food choice motives in the context of ‘a typical day’. Food choice motives will, therefore, probably result in stable segments. In line with this, test–retest reliability of the motives ranges between 0·71 and 0·83⁽ Reference Steptoe, Pollard and Wardle ^{89 )}. In addition, trade-off tendencies may result in uniquely responsive segments. Although most consumers want products that satisfy as many motives as possible⁽ Reference Haley ^{26 )}, consumers do differ in the relative importance of these motives⁽ Reference Onwezen, Reinders and Van Der Lans ^{92 )}. Among elderly, Locher et al. ⁽ Reference Locher, Ritchie and Roth ^{73 )} found a general interest in the motives convenience, sensory appeal and price, but considerable heterogeneity regarding the motives health, natural content, familiarity and weight control⁽ Reference Locher, Ritchie and Roth ^{73 ,} Reference De Almeida, Graça and Afonso ^{82 )}.

Consumers who are motivated by health are not necessarily willing to try functional foods. These foods are expected to taste badly and perceived to be unnatural, unfamiliar and expensive⁽ Reference Poulsen ^{18 ,} Reference Ares, Giménez and Gámbaro ^{57 ,} Reference Verbeke ^{78 ,} Reference Landström, Hursti and Magnusson ^{87 )}, thus the importance of sensory appeal, natural content, familiarity and price also play a role. In addition, consumers' reasons behind food choice vary across different types of functional food⁽ Reference Ares and Gámbaro ^{76 ,} Reference Aschemann-Witzel, Maroscheck and Hamm ^{96 )}. As a result, segments based on food choice motives have been found to differ in acceptance of some but not all functional foods⁽ Reference Ares and Gámbaro ^{76 )}. This variability limits the responsiveness of segments based on food choice motives.

Furthermore, food choice motives are inferred from responses on the Food Choice Questionnaire, rather than measured directly. The questionnaire is lengthy with its thirty-six items; thus segments based on food choice motives are moderately identifiable. However, there is no consensus regarding the number and type of segments that may be derived from the Food Choice Questionnaire, which increases the influence that marketers have on the substantiality of segments. Segments that are too large may be broken down and segments that are too small may be combined with other segments. Nonetheless, segment substantiality is probably not a problem because segments based on food choice motives are often substantial enough ( ≥  21 %)⁽ Reference Honkanen, Olsen and Myrland ^{67 ,} Reference Ares and Gámbaro ^{76 )}. In contrast, segments may not be linked to media profiles, because not all food choice motives are reliably related to specific demographics. The price motive, for example, has been related to age in some studies⁽ Reference Honkanen, Olsen and Myrland ^{67 ,} Reference Pohjanheimo and Sandell ^{97 )} but not in others⁽ Reference Steptoe, Pollard and Wardle ^{89 )}.

Summary

Food-level segmentation bases are generally responsive and actionable because they are closely related to actual food choice. They are moderately identifiable (because they are inferred from multi-item questionnaires) and provide little information on accessibility (because they are not related to demographics). In addition, their ability to predict acceptance of specific functional foods is limited because they operate on a general food level. For example, consumers who appreciate functional foods in general may not be interested in specific functional food formats.

Product attributes and benefits sought

Food choice motives (for example, ‘health’) are fulfilled by product benefits (for example, ‘control of blood pressure’) which arise from product attributes (for example, ‘added omega-3 fatty acids’)⁽ Reference Van Raaij and Verhallen ^{27 ,} Reference Surai and Sparks ^{99 )}. These three concepts are often used interchangeably⁽ Reference Onwezen, Reinders and Van Der Lans ^{92 ,} Reference Olsen, Prebensen and Larsen ^{100 )} because they are closely related to each other. However, food choice motives reflect more general consumer tendencies, whereas the attributes and benefits that consumer seek probably differ across foods. Therefore, we evaluate product attributes and benefits together as product-level segmentation bases.

According to Haley⁽ Reference Haley ^{26 )}, the fact that consumers seek different benefits is the main reason why market segments exist. In various marketing domains, elderly consumers have been segmented on the product attributes and benefits they seek⁽ Reference Horneman, Carter and Wei ^{101 ,} Reference Shoemaker ^{102 )} and this has also been done for the general consumer population in the functional food market⁽ Reference Herath, Cranfield and Henson ^{103 ,} Reference Hailu, Boecker and Henson ^{104 )}. Examples of resulting segments are: ‘receptive to functional foods’ (47 %) and ‘receptive to general foods for wellbeing’ (53 %)⁽ Reference Herath, Cranfield and Henson ^{103 )}, or ‘pill lovers’ (23 %), ‘yoghurt lovers’ (16 %) and ‘pill loathers’ (16 %)⁽ Reference Hailu, Boecker and Henson ^{104 )}. Moreover, one of the few researchers who discussed the segmentation of elderly consumers in the food market recommended using benefit segmentation⁽ Reference Ahmad ^{24 )}. Nonetheless, this recommendation was not evaluated in practice.

The three attributes of functional food that are most frequently discussed in the literature are: functional ingredients (for example, Ca), health claims (for example, ‘increases bone mineral density’) and the carrier product in which functional ingredients are placed (for example, orange juice). These attributes each affect functional food acceptance individually⁽ Reference Van Kleef, Van Trijp and Luning ^{105 )}. In addition, ingredients and carriers interact with each other⁽ Reference Siegrist, Stampfli and Kastenholz ^{19 ,} Reference Krutulyte, Grunert and Scholderer ^{106 )} and, as a result, some carrier–ingredient combinations are more appealing to consumers than others⁽ Reference Poulsen ^{18 ,} Reference Landström, Hursti and Magnusson ^{87 )}. Functional food acceptance is also affected by the way in which health claims are presented. For example, elderly prefer health claims that focus on disease prevention (for example, ‘reduces risk of cancer’) over health claims that focus on health promotion (for example, ‘increases energy level’)⁽ Reference Bogue and Ryan ^{70 )}. However, carrier type is found to affect acceptance most strongly⁽ Reference Siegrist, Stampfli and Kastenholz ^{19 ,} Reference Ares and Gámbaro ^{76 ,} Reference Krutulyte, Grunert and Scholderer ^{106 )}.

Besides food-specific attributes (for example, functional ingredients), functional foods may differ on a range of general product attributes, such as product price, package, brand and additional ingredients⁽ Reference Sparke and Menrad ^{107 –} Reference Ares, Besio and Giménez ^{109 )}. These attributes also contribute to the benefits that a product provides. Although individual benefits are appealing to many segments, it is the total configuration of benefits sought that differ between segments⁽ Reference Haley ^{26 ,} Reference Onwezen, Reinders and Van Der Lans ^{92 )} and result in unique responsiveness. Overall, elderly prefer healthy carriers to unhealthy ones, but considerable differences exist in their specific preferences. For example, elderly are found to overall like protein-enriched bread and dislike protein-enriched candy, but differ in their acceptance of protein-enriched meat, microwave meals and canned soup⁽ Reference Van Der Zanden, Van Kleef and De Wijk ^{60 )}.

Segments based on product attributes and benefits sought are actionable because they can be directly translated into marketing efforts⁽ Reference Ahmad ^{24 ,} Reference Haley ^{26 )} (for example, by highlighting certain benefits) as well as product design (for example, by using certain carrier–ingredient combinations). In contrast, identifiability of these segments is limited. Researchers generally advise against directly measuring attributes and benefits sought⁽ Reference Van Kleef, Van Trijp and Luning ^{105 )} because consumers often do not exactly know why they do or do not want to purchase certain products⁽ Reference Ulwick ^{110 ,} Reference Riquelme ^{111 )}. Instead, researchers usually let participants evaluate product formats, from which the attributes and benefits sought are later derived using a statistical method (for example, conjoint analysis).

Generally, segments based on attributes and benefits sought are substantial enough ( ≥  16 %)⁽ Reference Hailu, Boecker and Henson ^{104 ,} Reference Chung, Hong and Kim ^{112 ,} Reference Hsu and Lee ^{113 )} but moderately stable. Sought attributes and benefits are linked to the food choice motives that are most important at the moment of food choice. These motives, in turn, can vary across situations⁽ Reference Onwezen, Reinders and Van Der Lans ^{92 )} and over time⁽ Reference Calantone and Sawyer ^{114 )}. However, when attributes and benefits sought are measured for a specific context, segments may be stable. In line with this, test–retest reliabilities of attributes and benefits sought have been found to vary strongly, from the weak 0·20 to the strong 0·93⁽ Reference Calantone and Sawyer ^{114 ,} Reference Bottomley and Doyle ^{115 )}. In addition, products and attributes sought cannot always be directly related to specific demographics⁽ Reference Hailu, Boecker and Henson ^{104 )} and may, therefore, provide little information on the accessibility of segments.

Past purchase

Aside from which attributes and benefits consumers seek, it can be useful to look at what products consumers actually purchase. Research suggests that past behaviour is strongly predictive of future behaviour⁽ Reference Ouellette and Wood ^{116 )} and when behaviour is frequently performed, it may turn into a stable habit⁽ Reference Conner, Armitage, Shepherd and Raats ^{117 )}. This applies to food choice as well⁽ Reference Gillespie and Johnson-Askew ^{118 ,} Reference Peters ^{119 )}. Past purchase may be especially predictive of future purchase in elderly because they live relatively stable lives⁽ Reference Amireault, Godin and Vohl ^{120 )} and may thus have well-established patterns of consumption. Indeed, even olfactory losses in the elderly are not strongly related to changes in food choice⁽ Reference Koster ^{121 ,} Reference Kremer, Bult and Mojet ^{122 )}. Segments based on product purchase are thus probably stable.

Nonetheless, the relationship between past and future behaviour is not especially meaningful⁽ Reference Ouellette and Wood ^{116 )}. Past behaviour and future behaviour are merely correlated with each other because they are based on the same underlying beliefs and motivations⁽ Reference Ajzen and Berkowitz ^{123 )}. Behaviour itself does not provide much information on why consumers do or do not purchase certain products and actionability for communication is thus limited. How can one, for example, promote functional food purchase among consumers who do not yet purchase functional food? In contrast, actionability for design may be moderate. For example, consumers who already purchase certain types of functional food may be interested in similar types of functional food and consumers may also be approached with functional alternatives of the conventional foods that they purchase.

Information on past behaviour may, however, only be useful in a market that is relatively static and the functional food market, in contrast, is one driven by innovation⁽ Reference Bäckström, Pirttilä-Backman and Tuorila ^{124 )}. Product formats or marketing efforts from the past may therefore be a weak starting point in the commercialisation of functional foods. Furthermore, few segmentation studies use purchase behaviour in isolation, because it tends to result in limited, substantial and heterogeneous segments (for example, buyers and non-buyers of a product) that are probably not uniquely responsive to marketing efforts. Therefore, behaviour is sometimes measured in combination with other variables, such as purchase frequency⁽ Reference Barrena ^{125 )} or motivations⁽ Reference Sparke and Menrad ^{107 )} to establish more diverse segments. Yet more often, past behaviour is used to describe segments rather than as a basis for market segmentation⁽ Reference Shoemaker ^{102 ,} Reference Littrell, Paige and Song ^{126 )}.

Segments can be identified using a self-report measure of past purchase, but self-report measures often provide unreliable results⁽ Reference Macdiarmid and Blundell ^{127 )}. In addition, consumers may be unaware of functional food purchases because consumers are often unsure whether or not products are ‘functional’⁽ Reference Van Der Zanden, Van Kleef and De Wijk ^{60 ,} Reference Landström, Hursti and Magnusson ^{87 )}. A better source of information is the customer card because it objectively records what customers purchase⁽ Reference Ziliani and Bellini ^{128 )}. Moreover, it records purchases across various occasions (for example, week and weekend days), which provides a better indication of behavioural tendencies than a single purchase⁽ Reference Fishbein and Ajzen ^{129 )}. Furthermore, customer cards are usually linked with customer profiles that contain various demographics as well as store patronage information on which media profiles have been based. This information can thus be used to determine the accessibility of segments. However, irrespective of the measure used, past purchase can only be measured when the product of interest is already for sale in stores. This variable may thus be of limited use for newly developed functional foods.

Summary

Product-level segmentation bases are unique in the sense that they are able to predict acceptance of specific functional food formats, to some extent. Similar to food-level segmentation bases, segments of product-level bases are relatively strong regarding stability, substantiality and actionability for design. In contrast, segments are not always responsive and actionable for communication. Whereas segments based on product attributes and benefits sought are both actionable and responsive, this is not the case for past purchase. In addition, although past purchase may be the strongest predictor of functional food purchase among the segmentation bases in the present paper, it can exclusively be used for products that are already in stores.