This study was based on a comprehensive customer satisfaction survey, a multinational retailer administered to its customers from over 40 countries shopping in over 300 stores, as published in Bolton et al. (2022). The survey data contained 2.4 million responses to about 500 different questions, regarding recent interactions with the retailer, experiences across touchpoints, and customer demographics. Not all customers answered all the questions, being assigned randomly to various modules of the survey. The goal of our research was to investigate how touchpoints moderate the antecedents of customer satisfaction with service encounters by comparing online and in-store encounters. We used a hierarchical linear model (HLM) and found that online customers weigh cognitive and behavioral qualities more heavily, while they weigh emotional and sensorial qualities less heavily than in-store customers. Since firms have limited resources, this research narrows down on key qualities of service encounters, on how qualities should be standardized or customized in global omnichannel environments. Though HLM is an established technique in marketing and business research, using it on a large dataset proved challenging because the observations were nested across customers, stores, and countries, and not all customers interacted with the retailer across all touchpoints. These challenges, combined with the large number of respondents and identification of the relevant variables, increased the technical difficulties. In this methodological case study, we will focus on how to navigate technical difficulties and provide guidelines for researchers struggling with large datasets with nested data for which HLM is the most appropriate technique to analyze data.