Case Study: Gochujang
Materials and methods
Conjoint test
An online
conjoint survey was created using SSI Web (Sawtooth Software version
6.0.6) and deployed from 12 to 29 October 2007. Conjoint analysis refers
broadly to a group of experimental procedures through which the
structure of consumers' preferences can be estimated. The underlying
premise of the approach is that products can be regarded as a bundle of
attributes from which utility (i.e., the value or relative
attractiveness) is gained. The approach has been widely used in a
variety of food and beverage applications.
In this study, respondents were asked to complete an online questionnaire consisting of prescreening qualification questions, attitudinal question, and 12 CBC tasks, with four alternatives per task including a "None" option (Table 1).
Table 1. Attributes and Elements Used in Online Conjoint Survey.
| Brand | Blue Dragon, Encona, Linghams, Conimex, Wing Yip, Hot & Joy |
| Pack size | 150 mL, 220 mL, 280 mL |
| Container material | Glass, squeezable plastic |
| Country of origin | Korea, Thailand, China, UK/Europe, Malaysia |
| Flavor description | Hot chili sauce blended with garlic; mild chili sauce blended with garlic, sugar and lime; mild and slightly sweet chili sauce; hot and sweet chili sauce; hot chili sauce |
| Price | GBP 0.75, GBP 0.95, GBP 1.15, GBP 1.35, GBP 1.55 |
Based on prevailing UK market characteristics, the pack size/price combinations of a pack size of 220 mL at 75 p, and a pack size of 280 mL at 75 p or 95 p were not included in the experimental design.
To qualify to complete the survey, all respondents had to be aged 18–65 years, be responsible for the main grocery shopping of their household, consume and not reject a range of home-cooked oriental/far eastern cuisines, enjoy hot/spicy chili-based foods, and be willing to buy hot chili-based oriental sauces.
The cleaned datasets of 676 respondents were included in the analysis. This exceeds the recommended minimum sample size, based on the number of choice tasks, number of alternatives per task, and largest number of levels for any one attribute contained in this study.
Thirty-four percent of qualifying respondents were aged 18–34 years, 54% were 35–54 years, 10% were 55–65 years, and 74% of respondents were female. For the principal income earner of the household, 41% of respondents were of professional/managerial status, 25% were sales/administrative/technical workers, and 16% were skilled manual workers.
Respondents were
also asked to respond to 10 statements designed to gauge their reaction
to novel foods. The results illustrate that the survey respondents
were receptive to novel and ethnic foods.
Home placement test
Fifty-seven percent of respondents were aged 18–44 years, 62% were female, and 60% had an ABC1 socioeconomic profile. Similar to the conjoint study, respondents were also asked to respond to 10 statements designed to gauge their reaction to novel foods and again the results illustrated that the respondents were receptive to novel and ethnic foods.
In this test, respondents were instructed to evaluate two variants of the gochujang adapted sauce: one a barbecue-flavored sauce, and the second a chicken-bulgogi-flavored sauce. These samples were based on previous studies examining consumer acceptability of Gochujang-based sauces. The samples were presented in 290-g branded bottles, with one sample being placed with respondents in Week 1 and a second sample the following week; again evaluated over a 1-week period. In each evaluation period, the samples were assessed on three usage occasions. Respondents completed a pretrial questionnaire, corresponding to each use occasion and a post-trial questionnaire for each sample.
The sample presentation order was evenly alternated between respondents.
Statistical analysis
Conjoint test
A utility score represents the attractiveness of each feature in a conjoint study. The higher the utility score, the more that attribute or level influences consumer choices. In this study, individual utility scores were extracted using hierarchical Bayesian estimation and rescaled using a zero-centered differences method for easy comparisons.Only respondents with root likelihood values > 0.250 were included in the utility estimation. The root likelihood value is a prediction of respondent choices in a CBC and corresponds to the number of product concepts. In a survey with four product concepts, including the "None of these" option, consumer responses through the conjoint model should be predictable 25% of the time.
Importance scores were determined by calculating the utility score range of each attribute and dividing by the total utility range multiplied by 100. Latent class analysis and market simulator analysis were performed with Sawtooth Software SMRT version 4.16.0. A z test was undertaken on the share of preference values to identify significant variations in response to price changes.
One-way analysis of variance with a Tukey post hoc multiple comparison test on the zero-centered utility scores for the total population was applied using Minitab 17 Statistical software.