Making full use of qualitative data to generate new fish product ideas through co-creation with consumers: a methodological approach

Citation: López-Mas, L.; Claret, A.; Stancu, V.; Brunsø, K.; Peral, I.; Santa Cruz, E.; Krystallis, A.; Guerrero, L. Making Full Use of Qualitative Data to Generate New Fish Product Ideas through Co-Creation with Consumers: A Methodological Approach. Foods 2022,11, 2287. https://doi.org/10.3390/ foods11152287 Academic Editor: Cristina Calvo-Porral Received: 8 July 2022 Accepted: 27 July 2022 Published: 31 July 2022 Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations. Copyright: © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/). foods Article Making Full Use of Qualitative Data to Generate New Fish Product Ideas through Co-Creation with Consumers: A Methodological Approach Laura López-Mas 1,2 , Anna Claret 1, Violeta Stancu 3, Karen Brunsø 3, Irene Peral 4, Elena Santa Cruz 4, Athanasios Krystallis 5and Luis Guerrero 1,* 1Food Quality and Technology, Institute of Agrifood Research and Technology (IRTA), Finca Camps i Armet, s/n, 17121 Monells, Spain; [email protected] (L.L.-M.); [email protected] (A.C.) 2Department of Agri-Food Engineering and Biotechnology (DEAB), Baix Llobregat Campus, Universitat Politècnica de Catalunya (UPC), Building D4, st/Esteve Terradas, 8, 08860 Castelldefels, Spain 3MAPP Centre, Department of Management, Aarhus BSS, Aarhus University (AU), Fuglesangs Allé4, 8210 Aarhus, Denmark; [email protected] (V.S.); [email protected] (K.B.) 4AZTI, Food Research, Basque Research and Technology Alliance (BRTA), Parque Tecnológico de Bizkaia, Astondo Bidea, Edificio 609, 48160 Derio, Spain; [email protected] (I.P.); [email protected] (E.S.C.) 5Centre of Excellence in Food, Tourism and Leisure, American College of Greece (ACG), Gravias 6, 15342 Athens, Greece; [email protected] *Correspondence: lluis.guerr[email protected]; Tel.: +34-97-263-0052 (ext. 1494) Abstract: Co-creation is a process that directly involves different stakeholders in the idea generation phase of a new product development process. A pool of 112 new aquaculture fish product ideas was obtained by applying a combination of creative and projective techniques to the co-creation process with consumers in six focus groups conducted in three European countries (Germany, France, and Spain). The subjectivity of qualitative data analysis (e.g., focus groups) is one of its recognised disadvantages. To overcome this drawback, a combination of specialised software (i.e., Alceste), along with word frequency, co-occurrence, and context checking, was applied to provide a different approach to data analyses in qualitative studies. The method identified the most salient dimensions behind the participants’ discourse (naturalness, quality, ethics, price, and health) and inferred the importance those dimensions had for them, thus proving the existence of a correlation of 0.7 between what the participants said (frequency of mention) and what they liked the most (importance). Overall, the exploratory approach proposed is deemed useful for drawing key conclusions from qualitative research, thus offering an alternative to traditional content analysis. In future, the results obtained may be useful for selecting the co-created ideas with the greatest potential to be well received in the market. Keywords: focus group; new product development; aquaculture; seafood; content analysis; pseudo-triangulation; context; co-occurrence; word frequency 1. Introduction Approximately 80% of new food products launched in the market fail within the first year [ 1 ]. One of the most effective ways to enhance new product success in the market is to actively involve end-users during the new product development (NPD) process [2,3] . The early stages of NPD are crucial because failure at these stages is inexpensive compared with the cost of launching an unsuccessful product on the market [ 2 , 3 ]. Idea generation, screening, and the selection of the most promising ones are usually at the beginning of any NPD procedure [ 4 ]. Co-creation is a process that directly involves different stakeholders, such as consumers, in idea generation. According to Prahalad and Ramaswamy [ 5 ], co-creation can be defined as the “joint creation of value by the company and the customer, allowing the customer to co-construct the service experience to suit her context”. The collaboration Foods 2022,11, 2287. https://doi.org/10.3390/foods11152287 https://www.mdpi.com/journal/foods Foods 2022,11, 2287 2 of 18 between companies and consumers through co-creation not only allows the creation of customer value [ 6 , 7 ], but also examines consumers’ wants and needs [ 8 ], enhances NPD performance and reduces the risk of failing to meet consumer demands [ 9 ], reduces product development costs [ 10 ], and elicits more original and valuable ideas than those created by professional developers [ 11 ]. Co-creation can be applied in traditional creative techniques (e.g., brainstorming), in which the main goal is to generate a pool of ideas in a process that is primarily cognitive [ 12 ]. In contrast to creative techniques, projective techniques are based on the principle that people’s unconscious desires can be inferred by using ambiguous and unstructured stimuli, in which the subjects project their beliefs, attitudes, feelings, and motivations [ 13 ]. The more ambiguous and unstructured the stimuli, the more consumers may reveal their underlying personalities [ 3 , 13 ]. Banovi´c et al. [ 14 ] demonstrated that a combination of creative and projective techniques is a useful approach for generating new ideas in which consumers’ voices are widely incorporated. This combination of techniques can be applied to traditional qualitative methods, such as focus groups. Focus groups are often used to capture consumers’ spoken needs [ 8 ], gain insights into their behaviour, attitudes, needs, and wants, understand the underlying motives of their choices, generate new products, and accelerate NPD processes [ 15 , 16 ]. The traditional semantic analysis of focus group results involves the transcription of the discussion to draw main conclusions [ 17 ], a process that can be improved through the so-called “pseudotriangulation” [ 16 ]. Pseudo-triangulation is based on the independent analysis of three researchers or experts, reaching a consensus on a specific interpretation of the findings, which allows for the extraction of key conclusions [ 18 ]. Although the independent analysis of three researchers aims to balance the subjective influences of individuals [ 19 ], when interpreting qualitative data, it is common for a confirmatory bias to occur [ 20 ]; that is, researchers more strongly support the hypotheses that confirm what they believe [ 21 ]. Therefore, the subjectivity of the interpretation of results is one of the recognised disadvantages of qualitative techniques [22,23]. To overcome some of the limitations of traditional content analysis and manual coding, other alternatives can be used to increase the objectivity of the interpretation of results, such as the use of specialised software [ 24 ] (e.g., Alceste, ATLAS.ti, MAXQDA, NVivo). Software assistance is extremely useful during content analysis, as it allows for a reduction in content to take place in an objective way, saving time, cost, and effort [ 17 , 22 ]. In dealing with qualitative data, the less redundant the information there is, the easier the subsequent analysis is. Word frequency count and co-occurrent words are common outputs of qualitative data analysis software. Co-occurrence analysis reveals the relationship between words based on the number of times they are mentioned together [ 25 ], giving structure to the data. In contrast, word frequency counting is probably the most widespread softwareassisted method, as it is a simple and rapid way to summarise the results in terms of the most frequently mentioned words [ 26 ]. Hence, word frequency counting is commonly used in qualitative research as, for example, an indicator of the importance of certain words for participants [ 27 – 30 ]. Nevertheless, one of the main drawbacks associated with frequency counting is the loss of the context in which words are mentioned, which may lead to erroneous conclusions [ 17 , 31 , 32 ]. Therefore, because the meanings of words are often context-dependent, carrying out an in-depth qualitative analysis requires the interpretation of each word in its respective context [17]. In general, there is a lack of consensus on how to analyse and interpret qualitative data [ 17 ]. Most studies found in the literature seldom provide details on how they conducted the qualitative analysis or the complete set of codes applied [ 20 ]; some only mention “content analysis”. In addition, despite its advantages, the use of software for qualitative data analysis is still relatively limited in terms of studying the latent meaning of discourse [ 24 ]. Therefore, it could be hypothesised that the combination of software, along with the use of word frequency, co-occurrence, and context checking, could provide a different approach for data analyses in qualitative studies. The establishment of solid Foods 2022,11, 2287 3 of 18 methodological guidelines can be extremely useful when analysing qualitative data, as it has been criticised for its rather subjective nature for a long time. To prove the effectiveness of the methodology proposed in this study, the aquaculture fish sector was targeted. Food lifestyles are changing in Europe, and consumers usually have less time to spend on food preparation. New lifestyles, along with higher consumer awareness, have caused an increasing demand for a year-round supply of innovative and disruptive food products [ 33 ]. However, although the current food market seems to be saturated for most food categories [ 23 ], there are fewer new processed fish products in the market compared with other industries, such as meat [ 34 ], which leaves room for NPD, co-creation, and an exploration of the usefulness of qualitative data and its full potential. Accordingly, the aim of this exploratory study is threefold: (1) to propose an alternative approach to traditional qualitative content analysis by combining word frequency, co-occurrence and contextual analysis; (2) to explore the usefulness of the proposed approach in generating aquaculture fish product ideas and identify the most relevant product dimensions affecting potential acceptance by consumers; and (3) to explore whether word frequencies can be related to the subjacent relevant concepts or dimensions for the participants involved. 2. Materials and Methods 2.1. Participants’ Recruitment Purposive sampling with a predetermined quota for gender (evenly split) was used to select 36 participants in three countries (France, Germany, and Spain). A market research agency based in the three countries under study was subcontracted in June 2019 to recruit the participants and lead the moderation of the focus groups. Country selection was based on various aspects that may influence the generation of distinct ideas of fish products, including: differences in fish consumption per capita ( Spain > France > Germany ) [ 35 ]; differences in the main place of fish purchase—grocery store (Germany) or fishmonger/market (Spain) [ 36 ]; and the number of new fish products available in the national markets (France > Germany > Spain) [34]. In addition, participants also met the criteria of being older than 18 years, responsible for food purchase and preparation within their household, and fish consumers. In each country, fish consumption was used to divide participants into two groups: regular (at least once a week) and occasional fish consumers (three times a month or less), as it can be assumed that fish consumption frequency may be related to the demand for different fish products. 2.2. Focus Group Sessions Two face-to-face focus groups were conducted in each of the three countries targeted. Each focus group session with six participants lasted for two hours. The relatively low number of participants per focus group was selected based on the criteria that qualitative research does not intend to make inferences to a larger population, but to gain a deeper understanding of consumers’ perceptions and opinions about fish products [ 16 ]. Additionally, the qualitative approach used allowed participants to generate ideas about new aquaculture fish products. The moderators from the market research agency were previously briefed and followed a detailed discussion guide that included the following sections (Figure 1): 1. A warm-up debate about new foods and fish products. 2. A creative approach by applying direct analogies (also known as analogical thinking) [ 37 ] and storyboarding [ 14 ] to engage participants in generating new product ideas. Additionally, a reverse thinking task (also known as reverse brainstorming, tear-down, or purge) [ 38 ] was included to determine which characteristics of the ideas previously elicited would be rejected by the rest of the consumers. 3. A section where participants scored their acceptance of each new product idea elicited in the creative phase on a scale from 1 (“I very much dislike this fish product idea”) to 10 (“I like this fish product idea”). Foods 2022,11, 2287 4 of 18 4. A projective approach that included word association [ 13 , 14 , 18 , 39 ] and sentence completion tasks [ 13 , 39 ]. Four new fish product concepts identified by Gartzia et al. [ 40 ] were used as stimuli in both projective tasks to gain useful insights into consumers’ latent desires and feelings. 5. Finally, a general discussion section about fish and fish products to gain a deeper understanding of people’s practices regarding their purchase and consumption, opinions, interests, motivations, barriers, and behaviour with regard to new products. Foods 2022, 11, x FOR PEER REVIEW 5 of 20 Figure 1. Outline of the main steps taken during data analysis of the focus groups: (1) dimensions’ identification and their frequencies; (2) dimensions’ importance; and (3) dimensions’ importance versus dimensions’ frequency. 2.4.1. Part 1: Dimensions’ Identification and Their Frequencies The final corpus included the transcripts from all countries, arranged according to IMAGE [42], allowing the Alceste software, version 2018 (2018) (IMAGE, Toulouse, France) to perform content analysis in the following stages [24,41]: (1) text segmentation: into elementary context units (ECUs), i.e., basic analysable statistical units; (2) lemmatisation: simplifying words to their lemmas, i.e., to root forms that can be found in a dictionary (e.g., plurals into singulars); (3) reduction: discarding certain words (i.e., conjunctions, prepositions, pronouns); and (4) classification: the software’s internal dictionaries serve to sort the “content words”, i.e., words that were retained, by grammatical category (nouns, verbs, adjectives, and adverbs). An example of a quotation from one of the focus groups, “I buy new products to try out new flavours”, may be used to illustrate how Alceste reduces the content into two verbs (buy, try), one adverb twice (new), and two nouns (product, flavour), while all other components are discarded (I, to, out). Alceste tabulated some of the content analyses conducted; for example, a proximity matrix (co-occurrence within an ECU) and a contingency table (word frequency). Both outputs were modified according to the steps described in the next paragraphs, as illustrated in Figure 2. Figure 1. Outline of the main steps taken during data analysis of the focus groups: ( 1 ) dimensions’ identification and their frequencies; ( 2 ) dimensions’ importance; and ( 3 ) dimensions’ importance versus dimensions’ frequency. In addition, a brief discussion of salient ideas was carried out after each task to further understand the participants’ perceptions. The above-described focus group sessions were conducted in the native language of each country, audio-recorded and videotaped, simultaneously translated into English, and transcribed verbatim for further analysis from the English translation. 2.3. Transcript Preparation: Data Preparation Prior to data analysis, the transcripts were reviewed to detect and correct possible mistakes. Three researchers who were native French, German, or Spanish speakers listened to the recordings in their national languages and checked the English transcripts. Afterwards, all English transcripts were reviewed by the same researcher, following the guidelines proposed by Dalud-Vincent [ 41 ]. The spelling was standardised into British English (e.g., flavour instead of flavor) and the same spelling was used to designate the same concept (e.g., fillet instead of filet, the counterpart French word). The moderators’ speeches were removed from the transcripts, and only the participants’ individual speeches were retained. Finally, the content that was not relevant to the study was also removed (e.g., personal conversations between participants). 2.4. Data Analysis Data analysis was divided into three main parts: (1) dimensions’ identification and their frequencies; (2) dimensions’ importance; and (3) dimensions’ importance versus dimensions’ frequency. All steps taken are illustrated in Figure 1and are described in this section. Foods 2022,11, 2287 5 of 18 2.4.1. Part 1: Dimensions’ Identification and Their Frequencies The final corpus included the transcripts from all countries, arranged according to IMAGE [ 42 ], allowing the Alceste software, version 2018 (2018) (IMAGE, Toulouse, France) to perform content analysis in the following stages [ 24 , 41 ]: (1) text segmentation: into elementary context units (ECUs), i.e., basic analysable statistical units; (2) lemmatisation: simplifying words to their lemmas, i.e., to root forms that can be found in a dictionary (e.g., plurals into singulars); (3) reduction: discarding certain words (i.e., conjunctions, prepositions, pronouns); and (4) classification: the software’s internal dictionaries serve to sort the “content words”, i.e., words that were retained, by grammatical category (nouns, verbs, adjectives, and adverbs). An example of a quotation from one of the focus groups, “I buy new products to try out new flavours”, may be used to illustrate how Alceste reduces the content into two verbs (buy, try), one adverb twice (new), and two nouns (product, flavour), while all other components are discarded (I, to, out). Alceste tabulated some of the content analyses conducted; for example, a proximity matrix (co-occurrence within an ECU) and a contingency table (word frequency). Both outputs were modified according to the steps described in the next paragraphs, as illustrated in Figure 2. Foods 2022, 11, x FOR PEER REVIEW 6 of 20 Figure 2. Outline of the steps taken for the identification of the categories and dimensions and their frequencies. CA: correspondence analysis; MDS: multidimensional scaling. The original proximity matrix provided by Alceste had the same codes (words) in the rows and columns, while the cells were filled in with the frequency of co-occurrence (a measure of similarity), similar to a correlation matrix, but for categorical data [43]. Nevertheless, as Alceste does not consider the context in which words are said, a manual merging was conducted through the first pseudo-triangulation process with three independent researchers [18] to reduce the content. As a rule, throughout context checking in ECUs, when a word was used more than 75% of the time with a specific meaning, that meaning was attributed to that word. The reduction process through the first pseudo-triangulation consisted of the following stages: 1. Words with a common root (e.g., try, tried, and trying) were grouped under the same label (e.g., “try”). 2. Filler words, meaningless words, or sounds that consumers use while talking to fill in the pauses (e.g., basically, just, well) were removed after context checking. 3. Some words were grouped after checking their context in semantic (i.e., a set of words with related meanings) and associative semantic fields (i.e., similar to semantic fields but more subjective associations) to retain the information contained in the less frequently mentioned words. 4. Antonym words were grouped because they were considered to be extremes of the same scale (e.g., the “easy” label grouped the words easy, ease, difficult, and complicated). 5. Finally, homograph words, those with the same spelling but different meanings, were merged into their corresponding groups according to their meanings after context checking. Once the context was checked, the original proximity matrix provided by the Alceste software was reduced accordingly, thus grouping some words and removing others. Subsequently, a second reduction based on the frequency of mention was conducted, retaining only those words whose diagonal values in the proximity matrix were greater than 50. This cut-off was set by averaging all the values on the diagonal of the proximity matrix, rounding upward; therefore, only retaining words that co-occurred more frequently than the average. The words retained were labelled “categories”. The reduced proximity matrix was submitted to multidimensional scaling (MDS) analysis to plot the strength of the connections between co-occurring categories. Finally, some of the results from the MDS were used to group the categories into dimensions through a second pseudo-triangulation process. The selection of the final Figure 2. Outline of the steps taken for the identification of the categories and dimensions and their frequencies. CA: correspondence analysis; MDS: multidimensional scaling. The original proximity matrix provided by Alceste had the same codes (words) in the rows and columns, while the cells were filled in with the frequency of co-occurrence (a measure of similarity), similar to a correlation matrix, but for categorical data [43]. Nevertheless, as Alceste does not consider the context in which words are said, a manual merging was conducted through the first pseudo-triangulation process with three independent researchers [ 18 ] to reduce the content. As a rule, throughout context checking in ECUs, when a word was used more than 75% of the time with a specific meaning, that meaning was attributed to that word. The reduction process through the first pseudo-triangulation consisted of the following stages: 1. Words with a common root (e.g., try, tried, and trying) were grouped under the same label (e.g., “try”). 2. Filler words, meaningless words, or sounds that consumers use while talking to fill in the pauses (e.g., basically, just, well) were removed after context checking. 3. Some words were grouped after checking their context in semantic (i.e., a set of words with related meanings) and associative semantic fields (i.e., similar to semantic fields but more subjective associations) to retain the information contained in the less frequently mentioned words. 4. Antonym words were grouped because they were considered to be extremes of the same scale (e.g., the “easy” label grouped the words easy, ease, difficult, and complicated). Foods 2022,11, 2287 6 of 18 5. Finally, homograph words, those with the same spelling but different meanings, were merged into their corresponding groups according to their meanings after context checking. Once the context was checked, the original proximity matrix provided by the Alceste software was reduced accordingly, thus grouping some words and removing others. Subsequently, a second reduction based on the frequency of mention was conducted, retaining only those words whose diagonal values in the proximity matrix were greater than 50. This cut-off was set by averaging all the values on the diagonal of the proximity matrix, rounding upward; therefore, only retaining words that co-occurred more frequently than the average. The words retained were labelled “categories”. The reduced proximity matrix was submitted to multidimensional scaling (MDS) analysis to plot the strength of the connections between co-occurring categories. Finally, some of the results from the MDS were used to group the categories into dimensions through a second pseudo-triangulation process. The selection of the final dimensions was agreed upon by the three researches, who identified them based on the categories obtained, using input from the single-item food choice questionnaire (FCQ) [ 44 ]. The FCQ, originally developed by Steptoe et al. [ 45 ], is regarded as one of the most widespread methods used in consumer research to assess the motivations underlying food selection by measuring nine different factors. The original contingency table provided by Alceste with the word frequency count was reduced according to the process described in Figure 2. The words elicited (rows) were grouped according to the identified dimensions by simply adding the corresponding frequencies for each participant (columns). In addition, supplementary variables (columns) were added (i.e., country, focus group session, age category, gender, and fish consumption). A simple correspondence analysis (CA) was run to graphically display the reduced contingency table. 2.4.2. Part 2: Dimensions’ Importance The importance of dimensions was calculated by multiplying the acceptability given by participants with the scores given by a group of experts (see Table 1for an example). In detail, on one hand, the individual acceptability of the participants given to each idea generated using creative techniques (see Section 2.2) was used. On the other hand, a panel of eight experts from the food sectors of different European countries, with different backgrounds (i.e., academia and research centres) and expertise in NPD, evaluated all the ideas generated by the participants. These experts identified the dimensions (Table 2), from those obtained in the previous part (see Section 2.4.1), that were contained within each idea and their extent by scoring the dimensions identified in a continuous scale from 0 (“this dimension is not contained at all in this product idea”) to 10 (“this dimension is fully contained in this product idea”). Multiple factor analysis (MFA) was performed to determine the extent of agreement between the expert’s scores. As a high agreement was reached, the mean scores were used in subsequent analyses. The identification of the dimensions contained within each idea was necessary to explore whether there was a relationship between the importance of dimensions and their frequency of mention (see Section 2.4.3). In summary, the acceptability given by participants to each idea and the corresponding intensity for each dimension given by the experts were used to infer the importance attributed by the participants to each dimension. Foods 2022,11, 2287 7 of 18 Table 1. Fictitious example of the calculation of the dimensions’ importance for participant 1. Experts’ Mean Acceptability Dimension’s Importance Ideas Sensory Price Participant 1 Sensory Price Idea 1 2 6 9 18 * 54 Idea 2 7 4 5 35 20 ∑53 ** 74 * The importance of the sensory dimension for idea 1 was calculated by multiplying the experts’ mean scores for sensory dimension and the consumer’s (participant 1) acceptability for this idea. The same procedure was conducted for all dimensions, ideas, and participants. ** The dimension’s importance was then computed as the sum of all individual importances (for all the ideas); higher values in the sum indicates the higher importance of that dimension for participant 1. In this fictitious example, the dimension price is more important for participant 1 than sensory one. Table 2. Dimensions formulated by the grouping of the 44 categories through the second pseudotriangulation process. Dimension Categories Health Health, nutritious Process/preparation Product, preparation, condiments, cook, ingredient, recipe Sensory Colour, experience, flavour, gourmet, taste, try Quality Quality, fresh, origin, preservation Price Price, buy Familiarity Frequency, use Natural Natural, chemical Food_product Consumption, eat, fish, meat, food, meal Variety Variety, different, format, new, presentation, species Convenience Convenience, easy, snack Ethical Responsibility, packaging Occasion Occasion, people, share To illustrate which dimensions had a greater influence on participant acceptability, a preference map [ 46 ] was created for each focus group. For that purpose, the experts’ mean scores per dimension and the mean acceptability of participants per idea were used. In addition, to overcome the limitations linked to the fact that the participants only scored the ideas generated within their focus group, an overall preference map was created. The results of the preference maps could be useful when selecting those ideas to scale up in the NPD process. 2.4.3. Part 3: Dimensions’ Importance versus Dimensions’ Frequency To determine whether there is a relationship between what consumers say (dimensions’ frequency) and what they like most about the ideas generated (inferred dimensions’ importance), an MFA between these two matrices was conducted. The dimensions’ frequencies were gathered from the reduced contingency table used to conduct the CA (see Figure 2). The dimensions’ importance matrix was built using dimensions (rows) and the inferred dimension’s importance for each participant, as explained in Table 1(columns). All statistical analyses were performed using XLSTAT software, version 2020.1 (2020) (Addinsoft, Paris, France). 3. Results 3.1. Part 1: Dimensions’ Identification and Their Frequencies The automatic merging and reduction performed with the Alceste software allowed for a reduction in words from the 36,765 that had been in the initial corpus to only 747, retained for later analysis. After the first manual pseudo-triangulation process, 747 words were reduced to 186. Finally, from the 186 words obtained, only those whose diagonal values in the proximity matrix were greater than 50 were retained, as explained above, leaving 44 categories, as shown in the MDS plot (Figure 3). Foods 2022,11, 2287 8 of 18 Foods 2022, 11, x FOR PEER REVIEW 9 of 20 3. Results 3.1. Part 1: Dimensions’ Identification and Their Frequencies The automatic merging and reduction performed with the Alceste software allowed for a reduction in words from the 36,765 that had been in the initial corpus to only 747, retained for later analysis. After the first manual pseudo-triangulation process, 747 words were reduced to 186. Finally, from the 186 words obtained, only those whose diagonal values in the proximity matrix were greater than 50 were retained, as explained above, leaving 44 categories, as shown in the MDS plot (Figure 3). Figure 3. Multidimensional scaling of the 44 categories obtained using Alceste software, grouped by the first pseudo-triangulation process, and reduced by frequency of co-occurrence (Dimensions 1 and 2). The closer the categories are in the MDS plot, the higher the number of times those categories are mentioned together (co-occur). For example, the categories “ingredient” and “origin” (which grouped the word “local”) were repeatedly mentioned together, as exemplified in one quotation from a German participant: “Local ingredients are very important”. In addition, “quality” was frequently mentioned alongside the category “origin”, as shown in this quote by a Spanish participant: “It is the most determinant factor in the low quality, the origin of the fish”. As illustrated in another quotation from a focus group in France (i.e., “If it is quality fish, it is going to be expensive”), the categories “quality”, “fish”, and “price” (which grouped the word “expensive”) were frequently mentioned together. Interestingly, participants also frequently mentioned the categories “price” and “taste” in the same phrase; for example, “It is the price that decides, and the taste, of course” (a quote from a German participant). As previously explained, some of the results obtained from the MDS were used to group the 44 categories into 12 dimensions through a second pseudo-triangulation pro- Figure 3. Multidimensional scaling of the 44 categories obtained using Alceste software, grouped by the first pseudo-triangulation process, and reduced by frequency of co-occurrence (Dimensions 1 and 2). The closer the categories are in the MDS plot, the higher the number of times those categories are mentioned together (co-occur). For example, the categories “ingredient” and “origin” (which grouped the word “local”) were repeatedly mentioned together, as exemplified in one quotation from a German participant: “Local ingredients are very important”. In addition, “quality” was frequently mentioned alongside the category “origin”, as shown in this quote by a Spanish participant: “It is the most determinant factor in the low quality, the origin of the fish”. As illustrated in another quotation from a focus group in France (i.e., “If it is quality fish, it is going to be expensive”), the categories “quality”, “fish”, and “price” (which grouped the word “expensive”) were frequently mentioned together. Interestingly, participants also frequently mentioned the categories “price” and “taste” in the same phrase; for example, “It is the price that decides, and the taste, of course” (a quote from a German participant). As previously explained, some of the results obtained from the MDS were used to group the 44 categories into 12 dimensions through a second pseudo-triangulation process. For instance, the co-occurring categories “natural” and “chemical” in the MDS were grouped under the higher-order dimension “natural”. The final dimensions identified are listed in Table 2. The results of the CA performed on the 12 dimensions graphically displayed the contingency table with the dimensions’ frequency of mention and the supplementary variables (Figure 4), which allowed for the comparison between qualitative variables. Foods 2022,11, 2287 9 of 18 Foods 2022, 11, x FOR PEER REVIEW 10 of 20 cess. For instance, the co-occurring categories “natural” and “chemical” in the MDS were grouped under the higher-order dimension “natural”. The final dimensions identified are listed in Table 2. The results of the CA performed on the 12 dimensions graphically displayed the contingency table with the dimensions’ frequency of mention and the supplementary variables (Figure 4), which allowed for the comparison between qualitative variables. The German participants frequently mentioned the “food” and “health” dimensions. The dimension “ethical” was cited several times by both German and French participants. In the analysis, France was located closer to the “natural”, “occasion”, “convenience”, and “process/preparation” dimensions, but also to the “sensory”, “familiarity”, and “quality” ones, and all three were frequently cited by Spanish participants. Finally, the “price” and “variety” dimensions were also located closer to Spain. Figure 4. Correspondence analysis of the frequency of mention of the 12 dimensions and supplementary variables: high fish consumption (H); low fish consumption (L); black, German participants (Ge); yellow, Spanish participants (Sp); and green, French participants (Fr). Country totals are expressed in broader coloured points: light blue, dimensions; and light grey, supplementary variables. 3.2. Part 2: Dimensions’ Importance Creative techniques (direct analogies and storyboarding) encouraged participants from all three countries to generate a pool of 112 new aquaculture fish product ideas. The differences between countries were observed. The Spanish participants generated a higher number of ideas (45 ideas), followed by the German (42) and French (25) participants. Small differences were found in the proportion of ideas generated in the focus groups within the same country. A slight difference in the number of ideas generated between the focus group with participants with high (23) and low (19) fish consumption Figure 4. Correspondence analysis of the frequency of mention of the 12 dimensions and supplementary variables: high fish consumption (H); low fish consumption (L); black, German participants (Ge); yellow, Spanish participants (Sp); and green, French participants (Fr). Country totals are expressed in broader coloured points: light blue, dimensions; and light grey, supplementary variables. The German participants frequently mentioned the “food” and “health” dimensions. The dimension “ethical” was cited several times by both German and French participants. In the analysis, France was located closer to the “natural”, “occasion”, “convenience”, and “process/preparation” dimensions, but also to the “sensory”, “familiarity”, and “quality” ones, and all three were frequently cited by Spanish participants. Finally, the “price” and “variety” dimensions were also located closer to Spain. 3.2. Part 2: Dimensions’ Importance Creative techniques (direct analogies and storyboarding) encouraged participants from all three countries to generate a pool of 112 new aquaculture fish product ideas. The differences between countries were observed. The Spanish participants generated a higher number of ideas (45 ideas), followed by the German (42) and French (25) participants. Small differences were found in the proportion of ideas generated in the focus groups within the same country. A slight difference in the number of ideas generated between the focus group with participants with high (23) and low (19) fish consumption was found in Germany. The 15 ideas with higher participants’ mean acceptability per country are shown in Appendix A (Tables A1–A3). A high regression vector (RV) coefficient (0.7) between the expert scoring allowed for the use of their mean scores given to each dimension, together with the participants’ mean acceptability, to plot a preference map for each focus group. The RV coefficient measures the similarity between two matrices of quantitative variables or two configurations resulting from multivariate analysis [ 47 ]. The vector model was the best fit in all cases, with the Foods 2022,11, 2287 16 of 18 Table A2. Ranking of ideas with higher participant acceptability in Germany. Ranking New Product Idea Total Score 1 Average Score 2SD 3 1 New packaging for frozen fish (e.g., bag instead of using foil) 55.0 9.2 1.0 2 Recyclable packaging (second use) 52.0 8.7 2.0 3 High-grade convenience products, local, quick 51.0 8.5 2.1 4 Avoiding problems (e.g., storage life, no shell, no fishbone, etc.) 47.0 7.8 2.2 5 Product with healthy ingredients (e.g., for crumb coating) like protein bars 46.0 7.7 2.7 6 Combination of different fish tastes and textures 45.0 7.5 1.8 7 New fish convenience products 44.0 7.3 2.5 8 A whole fish (fillable, already filleted, etc.) 44.0 7.3 2.4 9 No-waste burger 44.0 7.3 2.5 10 Packaging not made of plastic or aluminium 41.0 6.8 3.9 11 Fish with more power/enriched/all necessary ingredients contained 37.0 6.2 3.3 12 Product containing only fish, no other animal products 36.0 6.0 3.5 13 Canned fish with new supplements (e.g., lentils) 35.0 5.8 2.6 14 Fish-to-go (patty for burger, etc.), for microwave 35.0 5.8 2.4 15 Fish that does not taste fishy 32.0 5.3 2.4 1 Sum of all participants’ acceptability within a focus group, scored on a scale from 1 (I very much dislike this fish product idea) to 10 (I like this fish product idea). Minimum score: 6, maximum: 60. 2 Participants’ mean acceptability of every idea: sum of all participants’ acceptability divided by the number of participants within each focus group (6). 3SD: standard deviation. Table A3. Ranking of ideas with higher participant acceptability in Spain. Ranking New Product Idea Total Score 1 Average Score 2SD 3 1With flavours, to mix with a small sauce bag (fine herbs, olive oil, pepper, mustard, garlic, and lemon) 57.0 9.5 0.8 2 Seabass, seabream, meagre in dices or crumbs, frozen, and boneless 56.0 9.3 1.2 3 Fish hamburgers 51.0 8.5 1.4 4 Tasty and boneless fish products 50.0 8.3 1.2 5 Coloured spaghetti surimi to decorate and add flavour 49.0 8.2 2.8 6 Fish tray (such as cheese tray), two fish species in the same tray, fresh 49.0 8.2 1.2 7 Fillets, cubes, smoked, in brine 49.0 8.2 1.2 8 Seabass fillets 48.0 8.0 2.0 9 New formats for children (stars, trapezoids, triangles) 48.0 8.0 2.5 10 Sushi with nice flavours 47.0 7.8 1.9 11 Fish with sauce preparation, tasty, high quality 47.0 7.8 1.7 12 Fish with flavour of other things (e.g., octopus + cooked potato, anchovies + chip potato) 44.0 7.3 3.3 13 Fish products with an extra healthy component 44.0 7.3 1.2 14 Small fish pieces with intense flavour 39.0 6.5 2.4 15 Different fish assortment, shellfish, different tastes 36.0 6.0 2.2 1 Sum of all participants’ acceptability within a focus group, scored on a scale from 1 (I very much dislike this fish product idea) to 10 (I like this fish product idea). Minimum score: 6, maximum: 60. 2 Participants’ mean acceptability of every idea: sum of all participants’ acceptability divided by the number of participants within each focus group (6). 3SD: standard deviation. References 1. Melgarejo, R.; Malek, K. Setting the Record Straight on Innovation Failure; Nielsen: New York, NY, USA, 2018. 2. Moon, H.; Johnson, J.L.; Mariadoss, B.J.; Cullen, J.B. Supplier and customer involvement in new product development stages: Implications for new product innovation outcomes. Int. J. Innov. Technol. Manag. 2018,15, 1850004. [CrossRef] 3. van Kleef, E.; van Trijp, H.C.M.; Luning, P. Consumer research in the early stages of new product development: A critical review of methods and techniques. Food Qual. Prefer. 2005,16, 181–201. [CrossRef] 4. Chang, W. The joint effects of customer participation in various new product development stages. Eur. Manag. J. 2019 ,37, 259–268. [CrossRef] 5. Prahalad, C.K.; Ramaswamy, V. Co-creation experiences: The next practice in value creation. J. Interact. Mark. 2004 ,18, 5–14. [CrossRef] Foods 2022,11, 2287 17 of 18 6. Galvagno, M.; Dalli, D. Theory of value co-creation: A systematic literature review. Manag. Serv. Qual. 2014 ,24, 643–683. [CrossRef] 7. Hoyer, W.D.; Chandy, R.; Dorotic, M.; Krafft, M.; Singh, S.S. Consumer cocreation in new product development. J. Serv. Res. 2010 , 13, 283–296. [CrossRef] 8. Witell, L.; Kristensson, P.; Gustafsson, A.; Löfgren, M. Idea generation: Customer co-creation versus traditional market research techniques. J. Serv. Manag. 2011,22, 140–159. [CrossRef] 9. Hsieh, L.-F.; Chen, S.K. Incorporating voice of the consumer: Does it really work? Ind. Manag. Data Syst. 2005 ,15, 769–785. [CrossRef] 10. Potts, J.; Hartley, J.; Banks, J.; Burgess, J.; Cobcroft, R.; Cunningham, S.; Montgomery, L. Consumer co-creation and situated creativity. Ind. Innov. 2008,15, 459–474. [CrossRef] 11. Kristensson, P.; Gustafsson, A.; Archer, T. Harnessing the creative potential among users. J. Prod. Innov. Manag. 2004 ,21, 4–14. [CrossRef] 12. Garfield, M.J.; Taylor, N.J.; Dennis, A.R.; Satzinger, J.W. Research report: Modifying paradigms—Individual differences, creativity techniques, and exposure to ideas in group idea generation. Inf. Syst. Res. 2001,12, 322–333. [CrossRef] 13. Donoghue, S. Projective techniques in consumer research. J. Fam. Ecol. Consum. Sci. 2000,28, 47–53. [CrossRef] 14. Banovi´c, M.; Krystallis, A.; Guerrero, L.; Reinders, M.J. Consumers as co-creators of new product ideas: An application of projective and creative research techniques. Food Res. Int. 2016,87, 211–223. [CrossRef] [PubMed] 15. Earle, M.; Earle, R. (Eds.) New product development: Systematic industrial technology. In Case Studies in Food Product Development; CRC Press: Boca Raton, FL, USA, 2008. 16. Guerrero, L.; Xicola, J. New approaches to focus groups. In Methods in Consumer Research, Volume 1: New Approaches to Classic Methods; Ares, G., Varela, P., Eds.; Woodhead Publishing: Duxford, UK, 2018; Volume 1, pp. 261–296. 17. Stewart, D.; Shamdasani, P.; Rook, D. Analyzing focus group data. In Focus Groups: Theory and Practice; SAGE Publications: Thousand Oaks, CA, USA, 2007; pp. 109–133. 18. Guerrero, L.; Claret, A.; Verbeke, W.; Enderli, G.; Zakowska-Biemans, S.; Vanhonacker, F.; Issanchou, S.; Sajdakowska, M.; Granli, B.S. ; Scalvedi, L.; et al. Perception of traditional food products in six European regions using free word association. Food Qual. Prefer. 2010,21, 225–233. [CrossRef] 19. Denzin, N.K. The Research Act: A Theoretical Introduction to Sociological Methods; Routledge: New York, NY, USA, 2017. 20. Pokorny, J.J.; Norman, A.; Zanesco, A.P.; Bauer-Wu, S.; Sahdra, B.K.; Saron, C.D. Network analysis for the visualization and analysis of qualitative data. Psychol. Methods 2018,23, 169–183. [CrossRef] 21. García, J.A.; Rodriguez-Sánchez, R.; Fdez-Valdivia, J. Authors and reviewers who suffer from confirmatory bias. Scientometrics 2016,109, 1377–1395. [CrossRef] 22. Guerrero, L.; Guàrdia, M.D.; Xicola, J.; Verbeke, W.; Vanhonacker, F.; Zakowska-Biemans, S.; Sajdakowska, M.; Sulmont-Rossé, C.; Issanchou, S.; Contel, M.; et al. Consumer-driven definition of traditional food products and innovation in traditional foods. A qualitative cross-cultural study. Appetite 2009,52, 345–354. [CrossRef] [PubMed] 23. Symmank, C. Extrinsic and intrinsic food product attributes in consumer and sensory research: Literature review and quantification of the findings. Manag. Rev. Q. 2019,69, 39–74. [CrossRef] 24. Illia, L.; Sonpar, K.; Bauer, M.W. Applying co-occurrence text analysis with ALCESTE to studies of impression management. Br. J. Manag. 2012,25, 352–372. [CrossRef] 25. Deng, S.; Xia, S. Mapping the interdisciplinarity in information behavior research: A quantitative study using diversity measure and co-occurrence analysis. Scientometrics 2020,124, 489–513. [CrossRef] 26. Puerta, P.; Laguna, L.; Vidal, L.; Ares, G.; Fiszman, S.; Tárrega, A. Co-occurrence networks of Twitter content after manual or automatic processing. A case-study on “gluten-free”. Food Qual. Prefer. 2020,86, 103993. [CrossRef] 27. Ares, G.; Deliza, R. Studying the influence of package shape and colour on consumer expectations of milk desserts using word association and conjoint analysis. Food Qual. Prefer. 2010,21, 930–937. [CrossRef] 28. Ares, G.; de Saldamando, L.; Giménez, A.; Claret, A.; Cunha, L.M.; Guerrero, L.; de Moura, A.P.; Oliveira, D.C.R.; Symoneaux, R.; Deliza, R. Consumers’ associations with wellbeing in a food-related context: A cross-cultural study. Food Qual. Prefer. 2015 ,40, 304–315. [CrossRef] 29. Melendrez-Ruiz, J.; Claret, A.; Chambaron, S.; Arvisenet, G.; Guerrero, L. Enhancing assessment of social representations by comparing groups with different cultural and demographic characteristics: A case study on pulses. Food Qual. Prefer. 2021 , 92, 104188. [CrossRef] 30. Polizer Rocha, Y.J.; Lapa-Guimarães, J.; de Noronha, R.L.F.; Trindade, M.A. Evaluation of consumers’ perception regarding frankfurter sausages with different healthiness attributes. J. Sens. Stud. 2018,33, e12468. [CrossRef] 31. Hsieh, H.F.; Shannon, S.E. Three approaches to qualitative content analysis. Qual. Health Res. 2005,15, 1277–1288. [CrossRef] 32. Vidal, L.; Ares, G.; Machín, L.; Jaeger, S.R. Using Twitter data for food-related consumer research: A case study on “what people say when tweeting about different eating situations”. Food Qual. Prefer. 2015,45, 58–69. [CrossRef] 33. Guiné, R.P.F.; Florença, S.G.; Barroca, M.J.; Anjos, O. The link between the consumer and the innovations in food product development. Foods 2020,9, 1317. [CrossRef] [PubMed] 34. Mintel. Global New Products Database (GNPD). Available online: https://www.gnpd.com/sinatra/search_results/?search_id= Fvf0PuANFb&page=0 (accessed on 2 December 2021). Foods 2022,11, 2287 18 of 18 35. EUMOFA. The EU Fish Market; European Commission: Luxembourg, 2021. 36. European Commission. Special Eurobarometer 515: EU Consumer Habits Regarding Fishery and Aquaculture Products; Maritime Affairs and Fisheries: Brussels, Belgium, 2021. 37. Dahl, D.W.; Moreau, P. The influence and value of analogical thinking during new product ideation. J. Mark. Res. 2002 ,39, 47–60. [CrossRef] 38. Souder, W.E.; Zeigler, R.W. Review of creativity and problem solving techniques. Res. Manag. 1977,20, 34–42. [CrossRef] 39. Vidal, L.; Ares, G.; Giménez, A. Projective techniques to uncover consumer perception: Application of three methodologies to ready-to-eat salads. Food Qual. Prefer. 2013,28, 1–7. [CrossRef] 40. Gartzia, I.; Peral, I.; Alfaro, B.; Riesco, S.; Cruz, E.S.; Krystallis, A.; Brunsø, K.; Stancu, V.; Claret, A.; Guerrero, L. Deliverable D5.1: Identification of Product and Market Requirements of Aquaculture Chain Stakeholders; MedAID Project: Derio, Spain, 2018. 41. Dalud-Vincent, M. Trial and critique of Alceste as a tool for analyzing semi-structured interviews in sociology. Lang. Société 2011 , 135, 9–28. [CrossRef] 42. IMAGE. Préparation du Corpus. Available online: https://www.image-zafar.com/images/formatage_alceste.pdf (accessed on 21 November 2021). 43. Bazeley, P. Qualitative Data Analysis: Practical Strategies; SAGE Publications: London, UK, 2013. 44. Onwezen, M.C.; Reinders, M.J.; Verain, M.C.D.; Snoek, H.M. The development of a single-item Food Choice Questionnaire. Food Qual. Prefer. 2019,71, 34–45. [CrossRef] 45. Steptoe, A.; Pollard, T.M.; Wardle, J. Development of a measure of the motives underlying the selection of food: The Food Choice Questionnaire. Appetite 1995,25, 267–284. [CrossRef] [PubMed] 46. Carroll, J.D. Individual differences and multidimensional scaling. In Multidimensional Scaling: Theory and Applications in the Behavioral Sciences; Shepard, R.N., Romney, A.K., Nerlove, S.B., Eds.; Seminar Press: New York, NY, USA, 1972; Volume 1, pp. 105–155. 47. Robert, P.; Escoufier, Y. A unifying tool for linear multivariate statistical methods: The RV- coefficient. Appl. Stat. 1976 ,25, 257. [CrossRef] 48. Claret, A.; Guerrero, L.; Aguirre, E.; Rincón, L.; Hernández, M.D.; Martínez, I.; Peleteiro, J.B.; Grau, A.; Rodríguez-Rodríguez, C. Consumer preferences for sea fish using conjoint analysis: Exploratory study of the importance of country of origin, obtaining method, storage conditions and purchasing price. Food Qual. Prefer. 2012,26, 259–266. [CrossRef] 49. Conte, F.; Passantino, A.; Longo, S.; Vosláˇrová, E. Consumers’ attitude towards fish meat. Ital. J. Food Saf. 2014 ,3, 178–181. [CrossRef] 50. Pieniak, Z.; Verbeke, W.; Scholderer, J. Health-related beliefs and consumer knowledge as determinants of fish consumption. J. Hum. Nutr. Diet. 2010,23, 480–488. [CrossRef] 51. McConnell, J.D. The price-quality relationship in an experimental setting. J. Mark. Res. 1968,5, 300. [CrossRef] 52. Kole, A.P.W.; Altintzoglou, T.; Schelvis-Smit, R.A.A.M.; Luten, J.B. The effects of different types of product information on the consumer product evaluation for fresh cod in real life settings. Food Qual. Prefer. 2009,20, 187–194. [CrossRef] 53. Claret, A.; Guerrero, L.; Ginés, R.; Grau, A.; Hernández, M.D.; Aguirre, E.; Peleteiro, J.B.; Fernández-Pato, C.; Rodríguez-Rodríguez, C. Consumer beliefs regarding farmed versus wild fish. Appetite 2014,79, 25–31. [CrossRef] [PubMed] 54. de Andrade, J.C.; de Aguiar Sobral, L.; Ares, G.; Deliza, R. Understanding consumers’ perception of lamb meat using free word association. Meat Sci. 2016,117, 68–74. [CrossRef] [PubMed] 55. European Commission. Facts and Figures on the Common Fisheries Policy: Basic Statistical Data; European Union: Luxembourg, 2020. 56. Fishbein, M.; Ajzen, I. Predicting and Changing Behavior: The Reasoned Action Approach; Taylor & Francis: New York, NY, USA, 2011. 57. Bleiel, J. Functional foods from the perspective of the consumer: How to make it a success? Int. Dairy J. 2010 ,20, 303–306. [CrossRef] 58. Klincewicz, K.; D˛ebska, K. Limits of co-creation: Evidence from a large-scale empirical study of food consumers. In Proceedings of the ESA Conference, Barcelona, Spain, 31 August–3 September 2021. 59. Guerrero, L.; Colomer, Y.; Guàrdia, M.D.; Xicola, J.; Clotet, R. Consumer attitude towards store brands. Food Qual. Prefer. 2000 ,11, 387–395. [CrossRef]