Data driven positioning analysis of music streaming platforms

Abstract

Traditional methods of assessing brand positioning through customer surveys often yield biased responses and limited insights. Leveraging state-of-the-art NLP algorithms allows for a more comprehensive understanding of sentiment and underlying topics within user reviews. By extracting meaning and emotions from large volumes of text data, this approach offers a more efficient, cost-effective, and less biased method of evaluating brand positioning compared to traditional approaches. Building upon previous research that used for sentiment detection and classification, this study goes a step further by constructing separate topic models for positive, neutral, and negative reviews. This in-depth analysis of the emotional aspects of customer reviews on music streaming platforms provides valuable insights that can guide brand positioning decisions. This thesis explores the analysis of user reviews on competing music streaming platforms using the suggested methodology. The study acknowledges the increasing demand for music streaming platforms, with Spotify holding the largest market share, followed by Apple Music, Amazon Music and YouTube Music. Distinct topic models are created for positive, neutral, and negative reviews. Positive reviews highlighted various aspects of the music application, such as features, functionality, playlists, customization, and emotional associations with music. Neutral reviews encompassed opinions on music consumption, app-related challenges, music subscription services, advertising, monetization, and offline use. Negative reviews discussed general music activities and technical difficulties. Further analysis involved calculating the scores of each platform for each topic. The results indicated platform-specific associations with different topics. Nextly, the results of sentiment analysis and topic modelling are checked for accuracy using statistical methods. One method is chi-square analysis, which examines the relationship between user sentiment and star ratings. This helps to confirm the sentiment analysis results and understand how users' sentiment align with their ratings. Another technique used for validation is correspondence analysis, which is applied to the contingency table of survey results. A survey is created to gather to opinions on each platform including an extra platform called ideal brand, and evaluates each platform for each topic. These validation methods ensure the reliability of the findings. By employing data-driven methods for brand positioning, this study presents a novel approach that is taking place of traditional methods commonly used in assessing brand positioning. This approach leverages the data to provide more comprehensive and subjective evaluation. This data-driven approach offers a new perspective and a more precise comprehension of customer perceptions.