Customers’ final purchase decisions for electronic products are understandably influenced by previous experiences, marketing messages such as price and promotion, and opinions from other consumers (Simonson and Rosen 2014). In particular, millions of product reviews are posted daily on online review boards or social media represent aggregate consumer preference data (Decker and Trusov 2010). Past studies analyzing online reviews or word-of-mouth (WOM) have focused more on the quantitative dimension of volume of WOM (or “how much people say”), but less on qualitative dimension of valence of WOM (or “what people say”) (Gopinath, Thomas, and Krishnamurthi 2014). However, recent studies have analyzed disaggregate-level UGC by performing text mining in addition to a general analysis of volume and valence of OUGC. Onishi and Manchanda (2012) investigate the relationship between movie sales and both TV advertising and blogs. Although the authors find that the volume and the valence of OUGC (i.e., blogs) are predictive of market outcomes, they retain only certain words (i.e., advertising, award, interesting, and viewed) that consumers would find useful, therefore having general predictive power for market outcomes. Gopinath, Thomas, and Krishnamurthi (2014) address the relationship between the content of online WOM, advertising, and brand performance of cell phones and find that the volume of OUGC does not have significant impact on sales, but only the valence of recommendation UGC has a direct impact. Liu, Singh, and Srinivasan (2015) find that both the volume and sentiments of Tweets do not outperform the information content of Tweets in predicting TV series ratings. Although these three papers have investigated the importance of qualitative UGC through text mining techniques, such studies have not accounted for the detailed dimensions of specific contents. For example, Onishi and Manchanda (2012) use only 4 words out of top 30 frequently cited words for their analysis, and Gopinath, Thomas, and Krishnamurthi (2014) classify the OUGC into three disaggregated dimensions (i.e., attribute, emotion, and recommendation) without further classifications of subcategories and valence of positivity and negativity. Liu, Singh, and Srinivasan (2015) mainly focus on positive and negative Tweet contents about TV shows, lacking further classification of functional and emotional dimensions. In contrast to these studies, this study aims to examine in-depth multidimensional aspects of the content of online reviews, i.e., qualitative UGC, and their impacts on product sales. In this process, we develop defensible measurements of UGC by executing a comprehensive empirical text analysis and evaluate the impact of measures of qualitative UGC relative to volume measure of quantitative UGC. Specifically, we analyze a large data set of UGC on the 350 most talked-about smartphone games from seven different genres (e.g., action, arcade, shooting, puzzle, role playing, simulation, and sports) over a 30 month period, August 2010 to February 2013. We utilize a theoretical framework that classifies qualitative UGC into two major perceptions of functional and emotional dimensions. Prior studies show that perceptions of both functional (cognitive) and emotional (affective) dimensions should be considered to investigate their effects on perceived user satisfaction (Coursaris and van Osch 2015) and online shopping behavior (Van der Heijden 2004). It is evident that both functional and emotional UGC influence consumers to purchase a focal product (Lovett, Peres, and Shachar 2013). The functional UGC relates to the positive and negative attributes and beliefs about a product, and the emotional UGC pertains to the feelings and emotions in response to product experience. As an example, consider one innovative car-racing mobile game which, although expensive, has 3D graphics and high level of complexity. After playing this game, consumers may express their feedback on this game online by describing it as well-made, unique, but sometimes fearful (because a high bill charge is expected from excessive playing time), and addictive (because they like the game too much to stop playing it). This type of online reviews contains different types of UGC: functional (e.g., quality, innovativeness) and emotional (e.g., fear). Another layer of our analysis involves the heterogeneity of impact on product sales across different qualitative UGCs. Specifically, we consider the effects of functional UGC on product sales across emotional contexts such as anger and happiness, in other words, a simultaneous association between functional UGC and emotional UGC. For example, although a consumer may be attracted by some reviews on the high quality graphics of a mobile game (functional UGC), she may hesitate to purchase this product because other reviews express their fear about high cost of purchasing virtual goods (emotional UGC). Accordingly, we expect the functional UGC’s effects on sales to be moderated (amplified or reduced) by emotional UGC. We accommodate such interaction effects in both aggregate and disaggregate models. To the best of our knowledge, this study is the first to empirically identify two dimensions of qualitative UGC (functional and emotional), and shed light on the effects of multidimensional UGC categories on sales. Our findings on the influence of qualitative UGC on product sales are quite different from the prevailing view that firms should pay attention more to the volume of UGC (Chevalier and Mayzlin 2006; Liu 2006) but little to the valence of UGC (Duan, Gu, and Whinston 2008; Godes and Mayzlin 2004; Liu 2006). Rather, our research is in line with recent three papers (Gopinath, Thomas, and Krishnamurthi 2014; Liu, Singh, and Srinivasan 2015; Onishi and Manchanda 2012) in terms of the importance of considering specific contents from a vast amount of text data. However, our paper provides two key contributions. First, we show that specific categories of qualitative online UGC such as functional and emotional variables can be used to predict product sales; this result will be of a high managerial relevance. Especially, traditional methods that use simple metrics such as volume and valence of UGC are less accurate than our method that employs a sophisticated, multidimensional content analysis. Second, the results offer guidance to firms in determining which specific UGC (quantitative or qualitative; functional or emotional; under what contexts) they should focus on for increasing the efficiency of their online marketing activities. Utilizing a large dataset of online reviews on 350 mobile games consisting of four million postings generated for thirty months, the authors identified 76 representative words to describe the functional and emotional UGC using text analysis and word classification. We combined the resulting UGC volumes with weekly sales, resulting in 1,835 observations for analysis with hierarchical Bayesian methods. We find that functional UGC includes 54 representative words to describe various levels of product quality, product innovativeness, price acceptability, and product simplicity, and emotional UGC includes 22 words to express anger, fear, shame, love, contentment, and happiness. The results show that the volume and valence of aggregated functional UGC and the share of aggregated emotional UGC have the positive effects on sales. The volume and valence of functional UGC subcategories have mixed effects on sales and the link is moderated by the share of emotional UGC subcategories. These results are in contrast to those in the literature. Further, a sales forecasting model which includes 13 variables of UGC subcategories shows the best predictive validity. The authors discuss the implications of these results for online marketers.