The brand's core image is not just formed by the things the company creates, including products, logos, packaging. There is also the aspect where the users influence the brand's core image. In this paper, we apply the analytical method in Shirai et al. (2016) to try understanding a “brand's core image” by identifying the characteristics of a brand's typical users. Two universities were selected as targets for an empirical analysis. This is because the relationship between the university and the students is a very suitable target for clearly grasping the relationship between the brand's core image and the user's core image. Furthermore, this method cleared the image of typical users, not general users.

In this study, we analyzed factors that influence mutual favorability in Japan and Korea. Japanese and Korean high-school students, university students, and adults were asked to answer a questionnaire consisting of various items considered to be factors that influence mutual favorability. The respondents number 1,278 Japanese and 908 Koreans. As a result of the analysis, nine factors were identified as statistically significant in forming mutual favorability. In particular, four factors, “area of interest”, “interest in people”, “experience of visiting”, and “information on the Internet” of each country are strongly related to the formation of mutual favorability. Although there are minor differences between age groups among both Japanese and Koreans, “liking” was influenced when respondents are interested in entertainment, leisure, lifestyle and people or when they find positive information on the Internet. However, some factors indicate differences such as, Korean sports in Japan, Japanese animation and games in Korea. On the other hand, if they had no interest in each other's country, people and language, the option “other than liking” was influenced among both Japanese and Korean respondents. The same response is seen among respondents with no experience of visiting the other country or when they find negative information about each other's country on the Internet.

In recent, the concept of clumpiness has received increasing attention as a novel method called RFMC in Customer Relationship Management (CRM). However, it has been not so long since clumpiness is proposed, there still remains difficulties. In this paper we firstly identify the problems of clumpiness in both statistical and practical fields, propose the improved indication of clumpiness and apply it to some of the most preferential problems for firms such as the detection of early churn in smartphone game market and the prediction of consumer's share-of-wallet (SOW) in grocery market to validate its usefulness in marketing analysis.

We attempt to clarify the mechanism of the amount of money of purchase by each customer in a department store in Nagoya city using the Bayesian regression model. Specifically, we use two types of explanatory variables, “Direct mail”, and “Events”, to research a relationship between the purchase amount and department store's strategy. Direct mail means the number of mails that the department store uses to inform their customers about new items, privileges of card holders, and storewide events. We also add “frequency coming to the store” as explanatory variables. Additionally, we adopt type-1 Tobit model to presume the optimal value for the dependent variable, if the dependent variable in the regression model is zero. Furthermore, we examine the relation between department strategy and the customer's attributes by incorporating customer's attributes into the hierarchical model.

A topic model is an analysis method that extracts latent topics from words of documents. Topic models are used in many fields nowadays, and they are used for marketing or medical data, where data of a subject are often repeatedly measured. Although topic models for longitudinal data have been already proposed, a topic model that can be applied to more diverse repeated measures data needs to be proposed. Then, we proposed a topic model for repeated measures data and evaluated its performance using real-world prescription drug data of a hospital.As a result, the proposed method showed superior performance to existing methods.