Motivations for user satisfaction of mobile fitness applications: An analysis of user experience based on online review comments

The study addresses how user experience (UX) elements reflected in online reviews determine user satisfaction with mobile fitness applications, an increasingly important sector within digital health, especially amid the COVID-19 pandemic. Prior work has largely relied on survey-based approaches (e.g., TAM, UTAUT2, ECT), leaving a gap in leveraging large-scale, naturally occurring review data. This research proposes using two natural language processing methods—bag-of-words and sentiment analysis—to extract UX dimensions from Google Play Store reviews and test their relationships with user satisfaction. The research questions are: RQ1: What factors determine users’ satisfaction with mobile fitness applications? RQ2: Can users’ satisfaction be examined using the UX elements in their online review comments?

UX is a key determinant of satisfaction and technology adoption, commonly conceptualized through usability, usefulness, and affection. Definitions emphasize ease/effectiveness (usability), performance enhancement (usefulness), and psychological/emotional responses to design (affection). Prior studies show UX factors shape intention and satisfaction across services (e.g., mobile reservations, wearables, e-learning). Online review data offer accessible, honest user perspectives for UX analysis. Gamification—using game elements like challenges and badges in non-game contexts—can enhance UX and continued use; prior findings show enjoyment/playfulness relates to satisfaction and adoption. However, few studies have assessed mobile fitness applications’ UX using online reviews or linked UX dimensions directly to satisfaction.

Study context and data: Reviews were collected from the Google US Play Store. For non-gamified fitness apps, the top five apps were identified via the keyword “fitness,” yielding 15,000 reviews with ratings. For gamified fitness apps, apps meeting gamification criteria (e.g., narrative/plots) were identified using the same keyword, yielding 12,824 reviews. Pre-processing removed reviews under five words, non-English content, emojis/emoticons; tokenization, lemmatization, and POS tagging were applied. After pre-processing, 7,913 non-gamified and 8,548 gamified reviews (total 16,461) remained. Study 1 (Bag-of-words): Dictionaries representing three UX components—usability, usefulness, and affection—were constructed based on validated prior work. For each review, the proportion of words matching each dictionary was computed (e.g., 3/10 words matching affection yields 0.3 affection level). Multiple linear regression examined the effects of these proportions on user satisfaction (1–5 star ratings), separately for non-gamified and gamified apps. Study 2 (Sentiment analysis with LIWC): Using LIWC categories, five UX-related constructs were operationalized: hedonic values (positive emotion), user burden (negative emotion), expectation confirmation (comparisons), pragmatic values (work, leisure, home), and social values (social words); each measured as a 0–1 proportion. Multiple linear regression assessed their relationships with satisfaction, again separately for non-gamified and gamified apps. Additional approaches: LDA topic modeling and K-means clustering were also applied to extract UX dimensions for comparison; their explanatory power for satisfaction was assessed via multiple regression.

Study 1 (Bag-of-words): Non-gamified apps (R2=0.158): Affection (M=11.25%, SD=13.83%; β=0.364, p<0.001) and usefulness (M=5.75%, SD=10.68%; β=0.175, p<0.001) positively predicted satisfaction; usability (M=5.21%, SD=9.05%; β=−0.025, p<0.05) had a small negative association. Gamified apps (R2=0.050): Affection (M=9.50%, SD=12.73%; β=0.177, p<0.001) positively predicted satisfaction; usability (M=3.65%, SD=7.22%; β=−0.114, p<0.001) negatively predicted satisfaction; usefulness (M=1.92%, SD=5.53%; β=−0.012, p=0.255) was not significant. Study 2 (Sentiment analysis with LIWC): Non-gamified apps (R2=0.149): Hedonic values (β=0.316, CR=40.579, p<0.001), user burden (β=−0.162, CR=−20.917, p<0.001), pragmatic values (β=0.030, CR=3.911, p<0.001), and social values (β=0.086, CR=11.154, p<0.001) significantly predicted satisfaction; expectation confirmation was not significant (β=0.004, CR=0.528, p=0.597). Gamified apps (R2=0.164): All five constructs were significant—hedonic values (β=0.278, CR=31.738, p<0.001), user burden (β=−0.220, CR=−26.347, p<0.001), expectation confirmation (β=0.033, CR=3.945, p<0.001), pragmatic values (β=0.019, CR=2.149, p<0.05), and social values (β=0.112, CR=13.278, p<0.001). Additional methods: LDA exhibited low explanatory power (non-gamified R2=0.046; gamified R2=0.074) and K-means likewise (non-gamified R2=0.111; gamified R2=0.053). Overall, affective/hedonic content is a strong positive predictor of satisfaction, while negative emotions (user burden) reduce satisfaction; usefulness mattered only in non-gamified apps.

Findings show that UX elements derived from online reviews meaningfully explain satisfaction with mobile fitness applications. Affective experiences (affection/hedonic values) are the most influential positive drivers across both app types, while negative emotional content (user burden) is a strong impediment. Usefulness relates to satisfaction for non-gamified apps but not for gamified ones, suggesting gamified contexts may rely more on affect and engagement than instrumental utility. Usability terms associated negatively in regressions, possibly reflecting that usability is often mentioned in problem contexts within reviews. Social and pragmatic values also contribute positively, and expectation confirmation matters in gamified contexts. These results support the feasibility of using NLP on user reviews to monitor and improve UX and satisfaction, highlighting the importance of emotional and experiential design in mobile fitness services, especially those employing gamification.

The study demonstrates that NLP-based analyses of online review comments can extract UX elements that predict user satisfaction with mobile fitness applications. Using bag-of-words and LIWC-based sentiment analysis, the work shows affection/hedonic values as primary positive drivers, with negative emotions reducing satisfaction; usefulness is significant only for non-gamified apps. The research provides practical guidance for developers and UX practitioners to prioritize affective experience and usability improvements and to leverage review analytics for UX monitoring. Future research should incorporate user demographics, broaden to multiple app stores and locales, and employ advanced machine/deep learning NLP methods to enhance UX dimension extraction and predictive performance.

Demographic characteristics of reviewers were unavailable, precluding subgroup analyses. The dataset was limited to the Google US Play Store, which may constrain generalizability across markets and platforms. The NLP techniques were relatively simple compared to newer machine/deep learning models; more advanced methods may yield improved UX extraction and predictive accuracy.