Mindset metrics, the measurement of consumers’ perceptions, attitudes, and intentions, have a long tradition in marketing, particularly in advertising and branding. Some of the most usual mindset metrics are brand awareness, brand image, personality traits, and attribute importance. Brand awareness and other mindset measures have the form of texts (bag of words). And, a natural methodology for analyzing these variables is topic modeling and the popular Latent Dirichlet allocation (LDA) model. The LDA methodology assumes that brands or concepts are represented by clusters of brands in consumers’ minds. This study proposes an extension/modification of the LDA model for brand awareness and other mindset variables that incorporate Bernoulli observations instead of the Multinomial specification present in the usual LDA specification. This extension is relevant since, unlike words in texts, brands and mindset concepts are not repeated within a document and have a dichotomous form, present or absent. The proposed model is applied to two brand awareness datasets. The results show significant gains in both managerial insights in analyzing brand clusters and consumers’ profiles.

心态指标是衡量消费者感知、态度和意图的指标，在​​营销领域有着悠久的传统，尤其是在广告和品牌推广方面。一些最常见的心态指标是品牌知名度、品牌形象、个性特征和属性重要性。品牌意识和其他心态措施具有文本（词袋）的形式。而且，分析这些变量的自然方法是主题建模和流行的潜在狄利克雷分配 (LDA) 模型。LDA 方法假设品牌或概念由消费者心目中的品牌集群表示。本研究提出了对 LDA 模型的扩展/修改，以用于品牌知名度和其他心态变量，这些变量包含伯努利观察，而不是通常 LDA 规范中存在的多项式规范。这个扩展是相关的，因为与文本中的单词不同，品牌和心态概念不会在文档中重复，并且具有二分形式，存在或不存在。所提出的模型应用于两个品牌知名度数据集。结果表明，在分析品牌集群和消费者概况方面的管理洞察力都有显着提高。