Nutrition labelling: Employing consumer segmentation to enhance usefulness

Abstract

Recent changes in nutrition labelling and increased consumer attention to nutrition and healthy foods have pressed food manufacturers to redesign their labelling, packaging and communication strategies. However, clearly identifying and targeting meaningful segments of food label, users must precede these marketing decisions. On the basis of their attitudinal and behavioural characteristics, the present study attempts to cluster food label users in Canada. A fuzzy clustering approach results in identifying three segments that differ in their understanding and use of nutrition labels, knowledge of nutrition and manipulation of quantitative information on nutrition. The results have interesting managerial implications for food marketers, retailers, nutritionists and public policymakers.

Appendix

Fuzzy clustering approach

Fuzzy clustering consists of classifying a set of consumers {C1, C2 … Cn} according to a finite number of criteria {crit1, crit2 …}. The evaluation values are considered to be fuzzy numbers. A fuzzy number ã is a fuzzy set defined on ℜ and characterized by its membership function:

Where,

• a _inf⩽a⩽a _sup

• the function l: ]−∞, a _inf[ → [0, 1] is increasing so that l(x)=0 for all x⩽β _inf, β _inf<a _inf.

• the function r: ]a _sup, +∞[ → [0, 1] is increasing so that r(x)=0 for all x⩾β _sup, β _sup>a _sup.

μ _a assigns to each element x∈ℜ a real number in the interval [0, 1] and μ _a (x) represents the membership degree of x in the set {x=a: x∈ℜ}. In this article, we characterize fuzzy numbers by triangular fuzzy membership function where a=a _sup=a _inf.

We use x(ij) to denote the evaluation of consumer Ci for the criterion j. The segmentation is therefore based on the concept of distance. First, the nearest consumer to the ideal group (consumers with the highest scores on all variables) and the nearest to the nadir (consumers with the lowest scores on all variables) were identified. Then, a voting scheme was designed to cluster the whole sample into two groups. Respondents who weakly belong to the two previous fuzzy sets (that is, ideal and nadir) are grouped into a third cluster (that is, the average group). As all responses in relation to any of the 13 criteria range from 1 to 5, the three clusters can be defined as follows:

• The CI surrounds the extreme ideal respondent who has the maximum evaluation on all the criteria (that is, CI=(5, 5, …, 5)).

• The CN surrounds the extreme nadir respondent who has the minimum evaluation on all the criteria (that is, CN=(1, 1, …, 1)).

• The CA surrounds the average respondent who has the average evaluation on all criteria (that is, CA=(3, 3, … 3)).

In contrast to the crisp clustering method, where respondent membership in a group can be either 0 or 1, consumers in the fuzzy clustering context may belong to different clusters to varying extents. Thus, a consumer may belong concurrently to the ideal, the average and the CNs but to different degrees. In such cases, we need to specify the degree of membership in each of the three clusters. The membership function is computed based on distance. The distance between Ci and Ck is calculated as follows:

Where, Ci and Ck are, respectively, consumer i and k. xij is the response of consumer i to the criteria j. xkj is the response of consumer k to the criteria j.

To determine the membership of a consumer to each of the three clusters, clusters’ boundaries should be defined. The boundary represents the nearest point to the centre of one of the three clusters with zero membership (0). On the basis of the scale used for all 13 criteria (that is, from 1: very poor/strongly disagree to 5: excellent/strongly agree), the centre of the CI has a fuzzy vector (5, 5, …, 5) and a membership value equal to one (1). A consumer C _FI who has the fuzzy vector (2, 2, …, 2) can be considered the nearest point to CI and not belonging to the CI. Thus, this consumers’ membership in the CI is equal to zero. Hence, the membership in the CI of any consumer x can be defined as follows:

For the CN, the vector (1, 1, …, 1) has a membership equal to one. A consumer C _FN with the fuzzy vector (4, 4, … 4) can be considered the nearest point to CN with a membership equal to 0. Then, the membership in the CN can be defined as follows:

For the CA, the vector (3, 3, …, 3) is the centre of the cluster with a membership value equal to one. This cluster would have two border points: the consumer CI with the fuzzy vector (5, 5, …, 5) and the consumer CN with the fuzzy vector (1, 1, … 1). Both can be considered the nearest points to CA with memberships equal to zero. Note that D (CI, CA)=D (CN, CA). Consequently, the membership function in the CA can be defined as follows:

By defining membership functions for the ideal, the average and the CNs, any consumer with an average score, on all 13 variables, ranging from one to five can be assigned to each of the three clusters with different degrees of membership, ranging from zero to one (see Figure A1).

Figure A1

 Fuzzy clustering of nutrition label users.