Methodological Strategies for the development of a Meat Probiotic Product based on Data Science using consumer studies (Trabajo presentado en Latin Food 2020 )
Sensory analysis plays a decisive role in determining whether the consumer will accept a new product to be launched into market. This is of importance, considering that more than 50% of the developed products fail in market during the first year of the product's life. Therefore, a more accurate analysis is required, and one tool to achieve this is the use of machine learning analysis in conjunction with conventional sensory statistical analysis techniques. This combination of information will allows us to make better decisions. Data science permits taking responses from consumers to group them and to profile users, as well as to analyze preferences. It detects attributes that need to be optimized to improve product acceptance. Therefore, in this study, a developed probiotic meat product was compared with five other products from leading commercial brands in the market of Aguascalientes, Mexico.
The chemical composition results show that this product is a good source of protein and fiber with low fat content. Showing that its chemical composition is 71.93%, 13.59%, 6.74%, 5.19% and 15.06% of moisture, protein, fat, ash and BEEFE.
The uniqueness of this research is based on the use of data science, which becomes an important part to integrate different data in order to analyze consumer information for providing a better understanding of the new designed food product from a sensory point of view. Therefore, the meat products were sensory analyzed by 72 consumers and different methods and methodologies related to sensory analysis were applied. Data was analyzed by different statistical and machine learning techniques. These were carried out using two different consumer or affective tests: just right (JAR), in addition to check all that apply (CATA) study. The analysis of both JAR and CATA established that the prototype is close to optimal according to the attributes studied. Moreover, the sensory attributes were considered to be positive.
The results of different analyses carried out on the different sensory tests indicated that the prototype developed showed good results that could make it competitive in the market. The landscape segmentation analysis (LSA) showed that the developed product is close to a zone considered as optimal by consumers, which is in accordance with other analyses such as the one developed by this working group known as LPL (Liking Product Landscape). In the latter, hedonic and JAR scales are used to profile consumers and, based on these, to determine consumer preferences and whether they are determined by consumer groups. In addition, attributes that need to be optimized to improve product acceptance can also be identified. The hierarchical clustering and Gaussian mixture models showed two segments of possible consumers according to socio-demographic data.
The above shows us that by carrying out this type of data analysis based on consumer sensory analysis we can obtain complementary information. On the other hand, by using Data Science we can establish that we can obtain more information. This makes this type of analysis methodology suitable to assure success of a new product in the market.
Keywords: Check-all that apply (CATA), Just about right (JAR), Landscape Segmentation Analysis (LSA), Liking Product Landscape (LPL), Data Science (DS).