Scholarly article on topic 'The consumer trail: Applying best-worst scaling to classical wine attributes'

The consumer trail: Applying best-worst scaling to classical wine attributes Academic research paper on "Agriculture, forestry, and fisheries"

Share paper
Academic journal
Wine Economics and Policy
OECD Field of science
{"Wine attributes" / "Consumer choice" / "Off-premise setting" / "Best-worst scaling" / "Wine consumer"}

Abstract of research paper on Agriculture, forestry, and fisheries, author of scientific article — Fernando Nunes, Teresa Madureira, José Vidal Oliveira, Helena Madureira

Abstract The main goal of this study is to gain a better understanding of the buying behavior of wine consumers in Portugal. More specifically, the study identifies extrinsic attributes that influence wine purchase choices in a retail store, crossing-tabulating the results with six classification variables. The authors use the best-worst scaling method with eighteen reference attributes for designing, implementing, and analyzing responses to a survey of 250 wine buyers. Results reveal that the most significant reference attribute is whether consumers had tasted the wine previously. These findings for Portugal are in accordance with what has been observed in other Western countries. The second most important attribute, region of origin, is also commonly identified in the literature as a significant attribute. The classification variables of age and gender help to explain the behavior of the majority attributes. Using a latent class analysis, the authors obtained a set of three segments that are representative of Portuguese wine consumers. The findings presented here have important implications for wineries and wine distributors in their efforts to know their consumers better in an off-premise context and thereby to maximize profit.

Academic research paper on topic "The consumer trail: Applying best-worst scaling to classical wine attributes"



Available online at


Wine Economics and Policy 5 (2016) 78-86

The consumer trail: Applying best-worst scaling to classical wine attributes

Fernando Nunesa, Teresa Madureiraa, José Vidal Oliveirab, Helena Madureirac

aINSTITUTO POLITÉCNICO DE VIANA DO CASTELO - Escola Superior Agrária Convento de Refóios, 4990-706 Refóios do Lima, Ponte de Lima, Portugal bINSTITUTO POLITÉCNICO DE LISBOA - Professor Jubilado da Escola Superior de Comunicagao Social, Campus de Benfica, 1549-014, Lisboa, Portugal cUNIVERSIDADE DO PORTO, Departamento de Geografia/CEGOT Via Panorámica s/n, 4150-564 Porto, Portugal

Received 1 February 2016; received in revised form 12 October 2016; accepted 20 October 2016 Available online 27 October 2016


The main goal of this study is to gain a better understanding of the buying behavior of wine consumers in Portugal. More specifically, the study identifies extrinsic attributes that influence wine purchase choices in a retail store, crossing-tabulating the results with six classification variables. The authors use the best-worst scaling method with eighteen reference attributes for designing, implementing, and analyzing responses to a survey of 250 wine buyers. Results reveal that the most significant reference attribute is whether consumers had tasted the wine previously. These findings for Portugal are in accordance with what has been observed in other Western countries. The second most important attribute, region of origin, is also commonly identified in the literature as a significant attribute. The classification variables of age and gender help to explain the behavior of the majority attributes. Using a latent class analysis, the authors obtained a set of three segments that are representative of Portuguese wine consumers. The findings presented here have important implications for wineries and wine distributors in their efforts to know their consumers better in an off-premise context and thereby to maximize profit.

© 2016 UniCeSV, University of Florence. Production and hosting by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (

Keywords: Wine attributes; Consumer choice; Off-premise setting; Best-worst scaling; Wine consumer

1. Introduction

Research on consumer behavior in retail contexts has assumed a special significance in recent years as globalization and a proliferation of brands and products, particularly in the food sector, have made the understanding of consumer choice and brand salience crucial to the success of firms (Dolbec and Chebat, 2013; Hamlin et al., 2012; Grisaffe and Nguyen, 2011; Waterlander et al., 2011; Thomson et al., 2005). Among food products, consumer behavior with respect to wine has been the focus of numerous empirical research studies, with 100 refereed articles published between 2004 and 2012 alone (Lockshin and Corsi, 2012). This increasing interest is in part due to the increasing cultural value that is associated with wine by urban consumers, for whom the choice of a wine has

E-mail addresses: (F. Nunes), (T. Madureira), (J.V. Oliveira), (H. Madureira).

Peer review under responsibility of UniCeSV, University of Florence.

become a complex and challenging task (Allen and Germov, 2010). The wine trade has for its part experienced increased production, the economic impact of which is significant. As matter of fact, in 2015, according OIV (2016), the world wine trade increased by 1.8% in terms of volume (to 104.3 million hectoliters) and by an even more impressive 10.6% (to 28.3 billion EUR) in terms of value compared with 2014. Owing to an excess supply of wine, however, the sector has been experiencing downward pressure on prices and a difficult environment for brand-building (Rannekleiv et al., 2012). Because the hundreds of referrals in the market and the emergence of new brands and sub-brands, choosing a wine in a retail store has become a complex and difficult task that many Western consumers find intimidating (Lockshin, 2003; Sutanonpaiboon and Atkin, 2012). The myriad of items on offer, similarity in format, price diversity, and the increasing gap between ordinary consumers' knowledge and the complexity of the choices before them, are all factors that wine buyers must face. As pointed out by Perrouty et al. (2006), those who wish to purchase wine may experience confusion owing to a

2212-9774/© 2016 UniCeSV, University of Florence. Production and hosting by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (

number of cues on the label, such as brand name, region, and grape variety. Moreover, because wine carries strong social connotations, it is often felt to be an important purchase decision, even if it is not always an expensive one (Barber et al., 2009). In order to protect themselves, consumers value a set of intrinsic and extrinsic attributes that are complex in nature and therefore difficult to identify in terms of their relevance to consumption; and because intrinsic attributes are difficult to assess prior to purchase, buying behavior is often based on extrinsic cues (Cohen, 2009). Several studies conducted in different geographical contexts have focused on such attributes as tasted the wine previously, matching to food, origin of the wine, I read about it, grape variety, someone recommended it, brand name; medals/awards, attractive front label, and alcohol level below 13% (Casini et al., 2009; Cohen, 2009; Goodman, 2009; Jaeger et al., 2009; Allen and Germov, 2010). A comparative review of the results of these studies reveals some inconsistencies among the rated attributes and thus points to the need for further research, since an explanation of how these factors influence consumers would be of great interest to producers, entrepreneurs, and marketers. There is a particular need for insight into the off-premise setting, since more often than not, as just discussed, consumers are faced with a large number of items from which to select.

Various studies have developed a number of methods for collecting data on wine consumers' preferences. Lockshin and Hall (2003) survey the large number of studies of consumer behavior regarding wine based on simple rating scales methods. Finn and Louviere (1992), on the other hand, have drawn attention to the limitations of methods based on simple ordering of attributes in terms of the difficulty of interpreting and validating new attributes and the impossibility of conducting comparisons among them. Goodman et al. (2005), Remaud and Lockshin (2009), and Casini et al. (2009) agree that one way to avoid the bias inherent in simple ordering methods is to apply methods based on discrete choice (scaling methods), which allow consumers to set the level of preference for a particular attribute.

Best-worst scaling (BWS), also known as Max-Diffs, derives from the discrete choice method (Finn and Louviere, 1992; Marley and Louviere, 2005) and has become a popular method for studying the importance of a particular issue to an individual or group relative to other issues (Burke et al., 2013). BWS was introduced by Finn and Louviere (1992), who used it to measure public concern about food safety, and it has since been applied in various contexts, including the social sciences, consumer behavior, and health care (Burke et al., 2013; Cohen, 2009; Dekhili et al., 2011; Flynn et al., 2010; Jones et al., 2013; Marti, 2012).

BWS has gained in popularity because it is thought to have greater discriminatory power than other scale measures (Sirieix et al., 2011) and to allow for better comparisons among countries and segments (Cohen and Neira, 2004). Rather than being asked to rate items one at a time, respondents are shown a predefined set of candidate items, from which are asked to choose one each that they consider the best and the worst (Finn and Louviere, 1992). Two main advantages are associated with

BWS methodology: first, it involves a fairly simple task for respondents, since it is less cognitively demanding to select extremes on a scale than to rank all items simultaneously (Burke et al., 2013; Erdem et al., 2012; Jones et al., 2013); second, it provides sufficient information to the researcher so that precise and comparable individual-level scales can be calculated (Burke et al., 2013; Jones et al., 2013; Louviere and Islam, 2008; Marti, 2012). Moreover, a growing body of literature has used the BWS to generate empirical findings that have contributed to a deeper understanding of the rationale behind wine purchases (Sirieix et al., 2011; Goodman et al., 2005; Remaud and Lockshin, 2009).

In the specific case of the Portuguese wine market, to our knowledge, no national study of wine consumers' preferences in a retail setting has yet been undertaken using BWS. This is a particularly significant gap given that Portugal is a mature wine market characterized by great product diversity. The present research accordingly explores Portuguese wine consumers' preferences in a retail environment. First, we seek to explore how eighteen extrinsic attributes affect consumer choice at the moment when a bottle of wine is purchased for a special meal. Second, we seek to identify correlations among such classification variables as age and income level and the focus attributes. Finally, we suggest, based on our findings, ways to diminish the heterogeneity of wine consumers by identifying segments with recognizable their main features.

2. Methods

2.1. Questionnaire design

A two-part questionnaire was developed for this study. The questionnaire was prefaced by an explanation that its purpose was to identify the most important attributes when choosing a wine in a store. The first part of the survey included six classification questions (variables): gender, age, place of residence, frequency of wine consumption, involvement in the wine sector, and income level. The second and main part of the survey was designed to measure the importance that the respondents attached to specific attributes of wine using the BWS method. Interviewees were asked to consider a hypothetical situation in which they needed to purchase a wine that they would offer to friends during a special dinner. Eighteen specific attributes of wine purchasing decisions were selected (Table 1). Fifteen of these were chosen based on previous studies (Loose and Lockshin, 2013; Goodman et al., 2005; Goodman, 2009; Casini et al., 2009; Jaeger et al., 2009; Cohen, 2009; Corduas et al., 2013; Madureira and Nunes, 2013), and the remaining three were new attributes introduced by the researchers (namely, attributes 16-18 in Table 1). In order to stress the importance of the attribute price, we decided to consider a price below €10, thus simplifying the buying decision process by lowering the level of risk (financial, social and/or emotional), as well as a price above €15, in order to evaluate the more complex decision process regarding wines with a higher purchasing risk. The third new attribute introduced, controlled denomination of origin (CDO), is

mentioned in the literature (Corduas et al. 2013; Mtimet and Albisu, 2006; Chiffoleau et al., 2006; Duarte et al., 2010) and is particularly relevant to the Portuguese wine sector, where there are 31 CDO/RO (region of origin) regions. The authors also attempt to clarify differences in consumers' perceptions regarding CDO and RO regions.

In this study, the eighteen attributes were combined into eighteen choice sets of four items each, and respondents were asked to select the best and worst attribute in each set, i.e., the most and least important attribute of the decision to purchase a specific wine. Four or five items per set are regarded as optimal for respondent evaluation, since a greater number could lead to respondent fatigue (Sawtooth Software Inc., 2013). The question sets were balanced in factor frequency, positional frequency, and orthogonality and therefore satisfy optimal design characteristics according to Sawtooth Software (2013). As a consequence, each attribute appears the same number of times across all choice sets, and each pair of attributes appears only once within each set. Multiple versions of the survey were generated in order to increase variation in the position and combination of attributes across respondents, thereby reducing any potential context bias (Sawtooth Software Inc., 2013).

Table 1

Wine attributes.

1 Tasted the wine previously

2 Recommended by friends/relatives

3 Grape variety or varieties

4 Region of origin

5 Brand name

6 Producer/winery name on the bottle

7 Medal/award, visible on the bottle

8 Suggested by experts in the media

9 Matching food

10 Information on back label

11 Promotional display in-store

12 Attractive front label

13 Alcohol level

14 Winemaker name on the bottle

15 Bottle characteristics (color and shape)

16 Price below 10 Euros

17 Price above 15 Euros

18 Controlled Denomination of Origin (CDO) of wine

2.2. Sampling method and survey administration

A snowball sampling strategy was adopted. Respondents were recruited based on their access to the web and willingness to collaborate. The survey links were first distributed through email. Our invitation to participate in the survey also asked participants to share it through mail or social media with family members, friends, and colleagues aged 25 years or older and living in Portugal. The survey was administered from July to August 2014.

In order to guarantee representativeness, we applied weighting factors to the initial sample and then calculated the outputs (Malhotra, 2004; Oliveira, 2012). To do so, we compared the initial sample with the ideal (representative) sample using the cross-classification variables gender x age x region. We obtained a x2 = 153.2317;df = 39;p — 0.0000, which indicated that the initial sample was not representative. Therefore, a K weighting factor was estimated for each cell and then applied to the initial sample frequency in order to obtain a representative sample (Table 2).

Finally, after comparing the initial sample with the representative sample, chi-square values were x2 — 1.149242;df — 39;p — 1.0000, which means the weighted sample is representative.

2.3. Data analysis

We used a best-worst scaling method as described in Cohen and Markowitz (2002). We began by computing best-worst raw scores for each respondent (individual B-W) for each wine attribute. The square B/W values were then transformed into a standardized ratio scale (0-100) so that the scale presents the standardized importance weights, with the sum of all items being 100. The assumption is that an item is chosen a particular percentage of times when presented together with other items (Sawtooth Software Inc., 2013).

To test the significance of each attribute and of the classification variables, new variables were created by grouping contiguous values of existing ones into five categories. In other words, we created categorical variables (rescaled scores bin) from continuous scale variables (rescaled scores).

Table 2

Weighted sample (age intervals).

Men Women Total

25/34 35/44 45/54 + 54 25/34 35/44 45/54 + 54

North, except Metro Porto 5 6 6 10 5 6 6 13 58

Metro Porto 3 3 3 6 3 4 3 7 32

Centre Region 5 5 5 12 5 6 6 15 59

Metro Lisbon 5 5 4 9 5 6 5 12 51

Pen. Setubal, Alentejo, Algarve 4 5 4 10 4 5 5 12 50

22 25 23 47 23 26 25 60 250

To test the significance of the classification variables for each attribute, the rescaled scores were grouped on a five-position scale and thereby categorized into five levels of importance. The continuous values from the minimum to maximum rescaled scores by attribute were converted into a range on the five-point scale. Four cut points (thresholds) were introduced, and the rescaled scores of each attribute were assigned to these five new classes. The intersection of these new ordinal variables (5-point scales) with the classification variables allows us to test whether rows (attributes) and columns (classification variables) are independent or are related by applying the chi-square test.

BWS data is very useful for segmentation analysis because information is gathered in the form of choice responses, which are quite amenable to such modern segmentation methods as LCA models (Cohen and Orme, 2004). Using the rescaled scores, a share of importance/preference index was constructed using Sawtooth software. This calculation is based on the assumption that all attributes have equal importance (Cohen and Neira, 2003).

3. Results

3.1. Respondent characteristics

From a total of 460 responses received, 250 complete responses were retained, the remainder (210) being excluded on grounds of incompleteness or a similar flaw. Table 3 presents the descriptive statistic of the sample population, which is statistically representative of the Portuguese population. Of the 250 respondents, most

Table 3

Characteristics of survey respondents.

Classification variables Modalities No. %

Gender Male 116 250 46.4

Female 134 53.6

Age 25-34 years old 45 250 18.0

35-44 years old 50 20.0

45-54 years old 48 19.2

55 years + 107 42.8

Locale of residence North. except Metro Oporto 58 250 23.2

Metro Oporto 32 12.8

Centre region 59 23.6

Metro Lisbon 51 20.4

Pen. Setubal, Alentejo, 50 20.0


Consumption frequency Once a week 91 250 36.4

Two to five times a week 90 36.0

Regularly during the 69 27.6

Involvement in wine sector Yes 88 250 35.2

No 162 64.8

Income level:net monthly salary Less than 1,500 107 250 42.8

in Euros 1500-2499 90 36

2500 + 50 20

No answer 3 1.2

were women aged 35-44; 107 reported a monthly net income less than 1500 Euros; all drank wine on a weekly basis, and 69 regularly during the week. Respondents were also distributed evenly across Portugal's various geographic areas.

3.2. Consumer preferred attributes

The analysis of the attributes can be presented in a variety of ways. According to Loose and Lockshin (2013), the standardized ratio scale is reliable, since any less important attribute can be interpreted as a ratio relative to the most important attribute (Table 4).

For convenience, data have been sorted (by decreasing level) and are graphically displayed (Fig. 1).

Of the attributes, tasted the wine previously is unmistakably the most valued (100.0) in the process of choosing a bottle of wine; region of origin (73.2) and recommended by friends/ relatives (71.3) are important as well. On the other hand, grape variety or varieties (41.9) and brand name (39.3) are moderately rated by Portuguese wine consumers. Finally, alcohol level (8.7) and bottle characteristics (color and shape) (6.5) are the items ranked lowest, indicating that these attributes are of little significance in the evaluation process that precedes the purchase of a bottle of wine.

3.3. Impact of classification variables over attributes

The intersection of new ordinal variables (five-point scales) with the classification variables allows us to test for independence between rows (attributes) and columns (classification variables) by applying the chi-square test. This statistical procedure evaluates the relationship between the dependent variable (attribute importance) and the independent (classification) variables. Table 5 presents the statistical significance of the relationships between the eighteen select attributes and the classification variables, in which a smaller p value corresponds to a greater probability of relatedness.

Significant correlations can be found between the first three driven preference attributes (tasted the wine previously, region of origin, and recommended by friends/relatives) and two classification variables (gender and age), while these three most valued attributes are not correlated with any of the remaining classification variables. We draw attention to the fact that the two attributes related to price have different correlations with classification variables; for while price below 10 Euros is correlated with age and level of involvement, price above 15 Euros is not correlated with any of the classification variables.

Neither winemaker name on the bottle nor information on back label have any correlation with classification variables. It should be noted that two of the classification variables, level of consumption and income level, have no significant relationship with any of the dependent variables (importance of the attribute). Further, gender and age can be a useful variable for segmenting the global wine market once they are correlated with the majority of attributes. Regarding income level, this apparent inconsistency could be an artifact of the premise of the study, in that the wine being selected is said to be for

Table 4

Raw best and worst, average best-worst, and standardized aggregated importance weights.

Item number

Times selected best

Times selected worst

(B-W)/n Sqrt (B/W)

Standardized ratio scale

Standardized importance weights (%)

Tasted the wine previously 2 627,3 47.1

Region of origin 5 519.1 72.8

Recommended by friends/relatives 3 421.7 62.3

Matching with food 10 420.1 166.8

Grape variety or varieties 4 285.5 122.3

Brand name 7 374.5 181.8

Controlled Denomination of 6 261.5 149.0 Origin (CDO)

Producer/winery name on the 17 264.3 174.6 bottle

Medal/award. visible on the bottle 8 270.3 223.4

Suggested by experts in the media 9 195.6 230.8

Information on back label 11 182.4 224.5

Price below 10 Euros 16 188.4 243.4

Price above 15 Euros 15 114.8 305.1

Winemaker name on the bottle 18 102.8 339.5

Promotional display in-store 12 98.1 408.6

Attractive front label 13 74.7 381.9

Alcohol level 14 53.3 533.6

Bottle characteristics (color and 1 35.1 621.8 shape)

2.321 1.785 1.437 1.013 0.653 0.771 0.450

- 0.141

- 0.169

- 0.220

- 0.761

- 0.947

- 1.242

- 1.229

- 1.921 2.347

3.65 2.67 2.60 1.59 1.53 1.44 1.32

1.10 0.92 0.90 0.88 0.61 0.55 0.49 0.44 0.32 0.24

71.3 43.5 41.9 39.3

24.7 24.1

16.8 15.1

13.4 12.1 8.7 6.5

6.4 5.9

4.0 3.9 2.7 2.4

2.2 2.0 1.4

Fig. 1. Standardized ratio scale relating to the eighteen wine attributes.

dinner with friends on a special occasion, potentially resulting in undervaluation by consumers.

3.4. Segmentation of consumers

LCA models are suitable for segmentation strategies, so a share of the importance/preference index was built by applying Sawtooth software to the rescaled scores. Table 6 summarizes the results.

Analyzing the results of the rescheduling, we find that the segments are not perfectly differentiated, in that the attributes tasted the wine previously, recommended by friends/relatives, and region of origin are regarded as important by respondents in all three segments. Nevertheless, when we concentrate on the attributes that present the higher score in each of the three segments (transversal reading), we find some interesting

defining characteristics. Thus Segment 1 is the most representative group, accounting for 44.9% of the sample and integrating respondents who assign greater relevance to the attributes matching food, grape variety or varieties and controlled denomination of origin (CDO) wine. This segment can be usefully described as the Expert Consumers group. Segment 2, representing 28.2% of the sample, is composed of respondents who attach the greatest importance to the attribute tasted the wine previously, followed by recommended by friends/relatives, brand name, suggested by experts in the media, price below 10 Euros, price above 15 Euros, attractive front label, promotional display instore, and bottle characteristics (color and shape). This segment may be designated the Influenced Consumers group. Finally, Segment 3 consists of 26.9% of the sample and includes subjects who care most about region of origin, medal/award visible on the bottle, producer/winery name on the bottle, information on back label, and winemaker name on the bottle, and may be called the Objective Cues Consumers group.

4. Discussion

This study has established a hierarchical relationship among extrinsic attributes that affect Portuguese consumers' choices at the moment of buying a bottle of wine for a special occasion. The results of best-worst analysis have been presented, along with a correlation between classification variables and extrinsic attributes of consumption and a proposed segmentation of consumers. The findings concerning consensus and mismatches in consumers' choices, as well as possible insights into marketing approaches, deserve further discussion.

By applying a BW standardized ratio scale, we have determined that tasted the wine previously, region of origin, and recommended

Table 5

Relationship between the dependent variable (attribute importance) and the independent (classification) variables (chi-square Test).


Classification variables Sign p values

Gender Age Residence region Level consumption Level involvement Level income

Bottle characteristics (color/shape) 0.28 0.35 0.05n 0.24 0.66 0.35

Tasted the wine previously 0.00n 0.00n 0.14 0.38 0.09 0.06

Recommended by friends/relatives 0.00n 0.00n 0.34 0.68 0.29 0.43

Grape variety or varieties 0.02n 0.00n 0.84 0,12 0.09 0.47

Region of origin 0.00n 0.00n 0.63 0.20 0.19 0.55

Controlled Denomination of Origin (CDO) wine 0.02n 0.00n 0.21 0.10 0.25 0.07

Brand name 0.01* 0.00n 0.16 0.12 0.25 0.47

Medal/award, visible in the bottle 0.04* 0.03n 0.50 0.34 0.46 0.67

Suggested by experts in the media 0.00n 0.00n 0.37 0.10 0.17 0.03n

Matching food 0.12 0,00* 0.81 0,02n 0,07 0.09

Information on back label 0.11 0.12 0.28 0.52 0.24 0.86

Promotional display in-store 0.20 0.04n 0.59 0.57 0.09 0.78

Attractive front label 0.02n 0.07 0.19 0.77 0.63 0.13

Alcohol level 0.56 0.03n 0.41 0.14 0.80 0.34

Price above 15 Euros 0.19 0.07 0.68 0.60 0.23 0.64

Price below 10 Euros 0.06 0.02n 0.54 0.90 0.03n 0.10

Producer/winery name on the bottle 0.00n 0.00n 0.42 0.74 0.04n 0.24

Winemaker name on the bottle 0.13 0.15 0.52 0.63 0.16 0.80

Significant relationships P < 0.05.

Table 6

Attributes in order of importance by segment.

Attributes Seg. 1 44.9% Seg. 2 28.2% Seg. 3 26.9%

Tasted the wine previously 12.4 15.2 10,0

Recommended by friends/relatives 8.0 12.0 8.6

Region of origin 10.2 9.2 13.8

Matching food 15.7 5.0 1.6

Brand name 6.4 9.1 6.5

Grape variety or varieties 9.5 3.3 7.1

Medal/award, visible in the bottle 4.4 4.0 10.0

Controlled Denomination of Origin (CDO) wine 8.3 3.1 7.1

Producer/winery name on the bottle 4.8 4.9 8,0

Suggested by experts in the media 3.1 7.1 3.9

Price below 10 Euros 3.9 6.8 2.2

Information on back label 3.3 3.2 7.7

Price above 15 Euros 2.8 4.8 1.4

Winemaker name on the bottle 2.3 1.1 6.4

Attractive front label 1.3 4.5 1.9

Promotional display in-store 1.5 3,0 2.5

Alcohol level 1.4 1.6 0.8

Bottle characteristics (color/shape) 0.7 2.1 0.5

Total 100 100 100

by friends/relatives are the three most relevant attributes regarding consumer choice. An initial conclusion is that these three attributes can be considered non-presence attributes, in the sense that determinants of consumers' cues are pre-defined, and thus do not depend on factual information available at the point of purchase (e.g., the bottle and store). These results are in line with those of Forbes et al. (2010), who emphasize the fact that the quality of a wine is unknown until the bottle is opened, so that extrinsic product attributes are frequently used by consumers as heuristic cues to indicate quality.

The consumption behavior of Portuguese wine consumers resembles that of other Western buyers in that tasted the wine previously is often mentioned as the most relevant attribute when buying a bottle of wine in an off-premise context (Batt and Dean, 2000; Casini et al., 2009; Goodman, 2009). This finding is also consistent with the conclusions of Goodman (2009) who, working in a cross-cultural context, found that consumers in eight of twelve countries considered the attribute tasted the wine previously the most important when choosing a wine in a store (Goodman, 2009). The obvious implication is that marketers should focus their efforts on capturing the consumer's state of mind outside the off-premise setting. Although this consideration can resemble the 'chicken and egg dilemma,' it is critical to recognize that the activities of dissemination and promotion prior to purchase, such as participation in wine shows and wine courses, can without doubt affect consumers' choice of wines.

Respondents in the present study considered the region of origin attribute the second most important cue when choosing a special wine in a store. This result is also consistent with recent research on the topic conducted in other Western countries (Remaud and Lockshin, 2009). Thus, for example, Batt and Dean (2000) report that the origin of a wine is the third most important cue; Atkin and Newton (2012) suggest region rather than merely appellations as a promotional strategy to promote wine; and Sutanonpaiboon and Atkin (2012) stress the significance of perceived value of origin as a decision heuristic regarding consumers' choices of wine. These results should have an impact on the promotional strategies that are developed by wineries, distributors, and retailers. In fact, considering that the region of origin is the first face-to-face attribute to be ranked, strategies for promoting brands should emphasize connections to a specific region, particularly

when the region is well known for its wine. It should, however, be noted that not all geo-referenced attributes are equally important. Indeed, the attribute Controlled Denomination of Origin (CDO) cue is ranked seventh, its relative importance but half that of region of origin (RO).

The significance of the attribute recommended by friends/ relatives reported here (third) is in contrast with results from similar studies of countries with more consolidated wine cultures, such as France (Goodman, 2009) and Italy (Casini et al., 2009). The contextualization of wine as a social product should accordingly be taken seriously by those who design marketing strategies. Indeed, the relatively high ranking of the recommended by friends/relatives attribute suggests that communication strategies should be developed for relatively more familiar social contexts.

The factor found to be fourth most influential in this study, matching food, was identified as the most important attribute in a study conducted in France (Goodman, 2009), and the second most important in another conducted in Italy (Casini et al., 2009). As alluded to above, these results are to be expected for such advanced gastronomic cultures, where the interconnection of wine with food is decisive and drives consumer preferences for both choices. It is, however, important to emphasize that this attribute is not usually regarded as crucial for marketing strategies aimed at selling wine in stores in the Portuguese domestic market. Largely influenced by frequent discussions of the subject in the media, consumers in most cases lack sufficient knowledge to make a suitable wine and food pairing. It is therefore important for wineries and stakeholders in off-premise settings to internalize these findings and develop appropriate information and awareness campaigns to help consumers select the right wine for the right food.

Although several studies report price to be one of the main attributes affecting purchasing decisions (Orth and Krska, 2002; Mtimet and Albisu, 2006; Lockshin and Corsi, 2012; Duarte et al., 2010), this study found that neither price above 15 Euros nor price below 10 Euros were relevant items. This apparent inconsistency might be explicable in terms of what turns out to be an overly simple approach that takes price to be an abstract value. By presenting respondents with two price levels (each associated with a simple or complex decision purchase process), however, the variable price becomes quantified. This move seems to promote greater price awareness by forcing consumers to respond in an objective manner. While it is difficult to contextualize this observation owing to the apparent lack of relevant literature on the subject, it should still be of particular interest for entrepreneurs and marketers. Indeed, if further research confirms these findings, it may be possible to conclude that Portuguese consumers are not particularly price-sensitive when selecting a wine for a special occasion.

Turning briefly to two other attributes, promotional display in-store ranks fifteenth, which indicates that promotional efforts may have an impact on consumers' likelihood of buying a specific bottle of wine. Marketers should thus perhaps expect low returns from static promotional activities in stores. Sixteenth is attractive.

Front label, a ranking that should be of particular interest to the large number of wine companies in Portugal that invest relatively large amounts in innovative labeling and might do better instead to rebuild their product and communication strategies.

Our analysis of the relationships among the attributes and the classification variables, then, reveals significant correlations between tasted the wine previously, region of origin, and recommended by friends/relatives and the classification variables age and gender. Moreover, these two classification variables are significantly correlated with the majority of the remaining attributes. These facts, together with the absence of significant correlations between the attributes and residence region, level of consumption, level of involvement, and level of income, highlight the importance of age and gender as segmentation variables. Further results are in line with those of a number of other researchers, including Chang et al. (2016), Thach and Olsen (2015), Forbes (2012), Atkin and Sutanonpaiboon (2007), and Barber et al. (2006).

As pointed out by McMillan (2013), the increasing level of segmentation of current consumers is helping wineries and distributors to become more efficient in their marketing efforts, thus increasing sales and capturing consumers with greater added value. Many companies in Portugal currently place their wines on the domestic market without considering the possibility of using segmentation strategies, probably because of the high level of heterogeneity that characterizes the domestic wine market and the limited research available on the subject. A significant number of brands and wineries thus arguably waste resources on inadequate strategies because they are unable to identify the needs and expectations of their various wine consumer segments. This study has identified three key segments, termed the Expert, Influenced, and Objective Cues Consumers groups. The Expert Consumers (44.9%), the most relevant group, value matching food issues, technical information about grape varieties, and information about Controlled Denomination of Origin (CDO). These consumers seek a deep understanding of enogastronomy and the identity of a wine. For members of this segment, wine is much more than just a drink in a bottle. This study draws attention to the relevance of CDO for expert consumers. Indeed, this segment can be identified by a set of specialized attributes indicative of highly conscientious wine-buying behavior.

The second segment, Influenced Consumers (28.2%), values a wide variety of mostly subjective attributes, such as sensory experiences and brand and bottle aesthetics. In addition, price below 10 Euros and price above 15 Euros are relevant. These consumers seem to seek attributes that in some manner diminish the purchasing risk and provide guarantee cues to access wine quality.

Members of the third segment, Objective Cues Consumers (26.9%), attach greater importance to such relatively more objective attributes as region of origin, medal/awards visible on the bottle, and producer/winery name on the bottle. These consumers still use purchasing cues, but ones that are more objective and targeted than those used by influenced consumers. Despite their insecurity, objective cues consumers show

a willingness to learn and prefer more detailed information, as marketers would do well to note.

By presenting three groups of wine consumers, this study aims to provide the wine industry a fresh perspective that can be incorporated into comprehensive marketing strategies that maximize a wine's competitive position. More research is needed to understand the Portuguese wine consumer better in both on- and off-premise settings. The present study is intended as a contribution to this deserving topic. Given the robustness of the BWS, the findings presented here are significant in their own right and as a foundation for future research.


'This work is supported by the European Regional Development Fund (ERDF) through the Competitiveness Operational Programme (COMPETE) [Project 006971 (UID/ SOC/04011) and by national funds provided by FCT, the Portuguese Foundation for Science and Technology, through its project UID/S0C/04011/2013'.


Allen, M., Germov, J., 2010. Judging taste and creating value: the cultural consecration of Australian wines. J. Sociol. 47 (1), 35-51.

Atkin, T., Newton, S., 2012. Consumer awareness and quality perceptions: a case for Sonoma country wines. J. Wine Res. 23 (2), 155-171.

Atkin, T., Sutanonpaiboon, J., 2007. A multinational study of gender wine preferences. In: Proceedings of the International DSI/Asia and Pacific DSI 2007. (

Barber, N., Almanza, B., Donovan, J., 2006. Motivational factors of gender, income and age on selecting a bottle of wine. Int. J. Wine Mark. 18 (3), 218-232.

Barber, N., Taylor, D., Dodd, T., 2009. The importance of wine bottle closures in retail purchase decisions of consumers. J. Hosp. Mark. Manag. 18 (4), 597-614.

Batt, P., Dean, A., 2000. Factors influencing the consumer's decision. Aust. New Zealand Wine Ind. J. Mark. Suppl. 15 (4), 34-41.

Burke, P., Schuck, S., Aubusson, P., Buchanan, J., Louviere, J., Prescott, A., 2013. Why do early career teachers choose to remain in the profession? The use of best-worst scaling to quantify key factors. Int. J. Educ. Res. 62, 259-268.

Casini, L., Corsi, A., Goodman, S., 2009. Consumer preferences of wine in Italy applying best-worst scaling. Int. J. Wine Bus. Res. 21 (1), 64-78.

Chang , K., Thach, L., Olsen, J., 2016. Wine and health perceptions: exploring the impact of gender, age and ethnicity on consumer perceptions of wine and health. Wine Econ. Policy. (Available online 15 September 2016, In Press, Accepted Manuscript).

Chiffoleau, Y., Laporte, J.-P., Touzard, J.-M., 2006, Tradition and Modernity in the New New World of Wine: Market innovation in Languedoc AOC wines, 3rt International Wine Business & Marketing Research Conference, July 6-8. Ecole Nationale Superieure Agronomique Montpellier, France.

Cohen, E., 2009. Applying best-worst scaling to wine marketing. Int. J. Wine Bus. Res. 21 (1), 8-23.

Cohen, S., Orme, B., 2004. What's your preference? Asking survey respondents about their preferences creates new scaling decisions. Mark. Res. 16 (2), 32-37.

Cohen, S. Markowitz, P., 2002. Renewing market segmentation: some new tools to correct old problems. In: ESOMAR Conference Proceedings. Amsterdam, The Netherlands. pp. 595-612.

Cohen, S., Neira, L., 2003. Measuring preference for product benefits across countries - overcoming scale usage bias with maximum difference scaling. In: ESOMAR Conference Proceedings. Amsterdam, The Netherlands.

Corduas, M., Cinquanta, L., Ievoli, C., 2013. The importance of wine attributes for purchase decisions: a study of Italian consumer's perception. Food Qual. Prefer. 28, 407-418.

Dekhili, S., Sirieix, L., Cohen, E., 2011. How consumers choose olive oil: the importance of origin cues. Food Qual. Prefer. 22, 757-762.

Dolbec, P., Chebat, J.C., 2013. The impact of a flagship vs. a brand store on brand attitude, brand attachment and brand equity. J. Retail. 89 (4), 460-466.

Duarte, F., Madeira, J., Barreira, M., 2010. Wine purchase and consumption in Portugal - an exploratory analysis of young adults' motives/attitudes and purchase attributes. Cienc. e Técnica Vitivinícola 25 (2), 63-73.

Erdem, S., Rigby, D., Wossink, A., 2012. Using best-worst scaling to explore perceptions of relative responsibility for ensuring food safety. Food Policy 37 (6), 661-670.

Finn, A., Louviere, J., 1992. Determining the appropriate response to evidence of public concern: the case of food safety. J. Public Policy Mark. 11 (1), 12-25.

Flynn, B., Huo, B., Zhao, X., 2010. The impact of supply chain integration on performance: a contingency and configuration approach. J. Oper. Manag. 28 (1), 58-71.

Forbes, S., 2012. The influence of gender on wine purchasing and consumption: an exploratory study across four nations. Int. J. Wine Bus. Res. 24 (2), 146-159.

Forbes, S., Cohen, D., Dean, D., 2010. Women and wine: analysis of this important market segment. In: Proceedings of the 5th International Academy of Wine Business Research Conference. Auckland (NZ).

Goodman, S., 2009. An international comparison of retail consumer wine choice. Int. J. Wine Bus. Res. 21 (1), 41-49.

Goodman, S., Lockshin, L., Cohen, E., 2005. Best-worst scaling: a simple method to determine drinks and wine style preferences. In: Proceedings of the 2nd International Wine Marketing Business Conference. Sonoma, California (USA).

Grisaffe, D., Nguyen, H., 2011. Antecedents of emotional attachment to brands. J. Bus. Res. 64, 1052-1059.

Hamlin, R., Lindsay, S., Insch, A., 2012. Retailer branding of consumer sales promotions. A major development in food marketing?. Appetite 58 (1), 256-264.

Jaeger, S., Danaher, P., Brodie, R., 2009. Wine purchase decisions and consumption behaviours: insights from a probability sample drawn in Auckland, New Zealand. Food Qual. Prefer. 20 (4), 312-319.

Jones, A., Jones, D., Edwards-Jones, G., Cross, P., 2013. Informing decision making in agricultural greenhouse gas mitigation policy: a best-worst scaling survey of expert and farmer opinion in the sheep industry. Environ. Sci. Policy 29, 46-56.

Lockshin, L., 2003. Consumer Purchasing Behavior for Wine: What We Know and Where We Are Going. Marchés et Marketing du Vin. 57. Bordeaux École de Management.

Lockshin, L., Hall, J., 2003. Consumer Purchasing Behaviour for Wine: What we Know and Where We are Going. International Wine Marketing Colloquium, Adelaide, Australia.

Lockshin, L., Corsi, A., 2012. Consumer behaviour for wine 2.0: a review since 2003 and future directions. Wine Econ. Policy (1), 2-23.

Loose, S., Lockshin, L., 2013. Testing the robustness of best worst scaling for cross-national segmentation with different numbers of choice sets. Food Qual. Prefer. 27 (2), 230-242.

Louviere, J., Islam, T., 2008. A comparison of importance weights and willingness-to-pay measures derived from choice-based conjoint, constant sum scales and best-worst scaling. J. Bus. Res. 61 (9), 903-911.

Madureira, T., Nunes, F., 2013. Relevant attributes of Portuguese wines: matching regions and consumer's involvement level. Int. J. Wine Bus. Res. 25 (1), 75-86.

Malhotra, N., 2004. Marketing Research - An Applied Orientation. Prentice Hall, New Jersey.

Marley, A., Louviere, J., 2005. Some probabilistic models of best, worst, and best-worst choices. J. Math. Psychol. 49 (6), 464-480.

Marti, J., 2012. A best-worst scaling survey of adolescents' level of concern for health and non-health consequences of smoking. Social. Sci. Med. 75, 87-97.

McMillan, R., 2013. State of the Wine Industry 2013. Silicon Valley Bank. EVP and Founder, Wine Division.

Mtimet, N., Albisu, L., 2006. Spanish wine consumer behavior: a choice experiment approach. Agribusiness 22 (3), 343-362.

OIV - Organization Internationale de la Vigne et du Vin Global state of conditions report: developments and trends. Press release. 2016. pp. 1-3.

Oliveira, J., 2012. Marketing Research - Volume I. Silabo, Lisboa.

Orth, U., Krska, P., 2002. Quality signals in wine marketing: the role of exhibition awards. Int. Food Agribus. Manag. Rev. 4 (4), 385-397.

Perrouty, J.P., d'Hauteville, F., Lockshin, L., 2006. The influence of wine attributes on region of origin equity: an analysis of the mediating effect of consumer's perceived expertise. Agribusiness 22 (3), 323-341.

Rannekleiv, S., Soccio, M., Mutis, V., 2012. Rabobank Wine Quarterly Q4. Quarterly report. Rabobank International.

Remaud, H., Lockshin, L., 2009. Building brand salience for commodity-based wine regions. Int. J. Wine Bus. Res. 21 (1), 79-92.

Sirieix, L., Remaud, H., Lockshin, L., Thach, L., Lease, T., 2011. Determinants of restaurant's owners/managers selection of wines to be offered on the wine list. Int. J. Retail. Consum. Serv. 18 (6), 500-508.

Sutanonpaiboon, J., Atkin, T., 2012. Using region to market wine to international consumers. J. Food Prod. Mark. 18 (1), 1-18.

Thach, L., Olsen, J., 2015. Profiling the high frequency wine consumer by price segmentation in the US market. Wine Econ. Policy 4 (1), 53-59.

Thomson, M., MacInnis, D., Park, C., 2005. The ties that bind: measuring the strength of consumer's emotional attachments to brands. J. Consum. Psychol. 15 (1), 77-91.

Waterlander, W., Scarpa, M., Lentz, D., Steenhuis, I., 2011. The virtual supermarket: an innovative research tool to study consumer food purchasing behaviour. BMC Public Health 11 (1), 589-599.