The use of a fitness app for customer recommendation: linear models and qualitative comparative analysis

The study addresses how fitness centre mobile applications influence customers’ future recommendations of the app. With the growing adoption of technological innovations in sports services, apps facilitate scheduling, progress tracking, and communication, potentially shaping user behaviour and loyalty. The research aims to determine customers’ evaluations of a fitness chain’s app and their impact on recommendation intentions, employing both linear models and fuzzy set qualitative comparative analysis (fsQCA) to capture direct effects and configurations of conditions. The research questions are: (RQ1) What factors influence future recommendations of mobile applications used in fitness centres? (RQ2) Do customers’ age and gender influence recommendations for using a sports service’s mobile app? The study’s importance lies in guiding managers on features that enhance two-way communication and user value to increase positive recommendations in a competitive fitness market.

The review highlights technological advancements in sports services, particularly wearables (e.g., smartwatches, GPS) that support performance monitoring, workload control, and injury prevention, as well as the role of social networks in engagement and marketing. In fitness centres, wearables enable personalized training and social interaction, supporting motivation and collaboration. The health benefits of physical activity are well-established, underscoring the value of technologies that facilitate activity tracking and guidance. In mHealth, evaluating app quality is crucial yet challenging; the Mobile App Rating Scale (MARS) provides a standardized framework across four dimensions: engagement, functionality, aesthetics, and information. Prior studies show fitness apps can promote physical activity, adapt to changing contexts (e.g., COVID-19), and that perceived app quality influences use intention. However, app quality evaluations often suffer from subjective measures and lack standardization. Theoretical framing uses MARS dimensions as predictors of recommendation, given their links to usability, reliability, and perceived value. Age and gender may moderate usage and recommendation patterns, with some evidence of differences by demographics and usage contexts.

Design and context: Quantitative study using hierarchical regression and fuzzy set qualitative comparative analysis (fsQCA) to assess how perceived app quality dimensions and user characteristics relate to recommendation of a fitness app (Fitbe, a university spin-off app integrated with fitness centre management software). Participants: N = 210 fitness centre users (54 females, 156 males). Age distribution concentrated between approximately 18–64 years (with female participants 26–57 reported). Customer permanence (fidelity) in the service categorized as ≤12 months, 13–36 months, and ≥37 months. Instrument: Spanish-validated MARS scale (Martin-Payo et al., 2021) assessing app quality across four dimensions with 16 items: Engagement (5 items), Functionality (4), Aesthetics (3), Information (4). Responses on 5-point Likert scales. Dependent variable: intention/recommendation of the app (1–5 agreement). Procedure: Researchers contacted fitness centres using Fitbe; eight centres participated. The app supports training tracking, coach communication, and activity booking. The online questionnaire took ~7–8 minutes. Data analysis: SPSS v25 used for descriptive statistics and hierarchical regression. Step 1 predictors: MARS dimensions (Engagement, Functionality, Aesthetics, Information). Step 2 added sociodemographics (age, gender) and permanence. fsQCA 2.0 used to perform calibration (with percentile anchors, e.g., 10/50/90), necessity, and sufficiency analyses. Model fit in fsQCA assessed via consistency (fit) and coverage (explained variance). Intermediate solutions reported; equifinality explored for high and low recommendation outcomes.

Hierarchical regression: - Step 1 (MARS dimensions predicting recommendation): Adjusted R² = 0.52 (p < 0.001). Significant predictors: Aesthetics (β = 0.36, p < 0.001) and Information (β = 0.30, p < 0.001). Functionality showed a positive but marginal effect (β = 0.13, p = 0.075). Engagement was not significant (β = 0.01, p = 0.89). - Step 2 (adding age, gender, permanence): No meaningful increase in explained variance (ΔR² = 0.01, p = 0.611); sociodemographics and permanence were not significant. fsQCA: - Necessity analysis: No single condition (Engagement, Functionality, Aesthetics, Information, Age, Gender, Permanence) reached consistency ≥ 0.90; thus, no necessary conditions for high or low recommendation. - Sufficiency (intermediate solutions): Multiple causal configurations lead to high or low recommendation (equifinality). • High recommendation: Best-explaining path included being female combined with high Engagement; Information appeared in two of three high-recommendation configurations. Total solution consistency ≈ 0.86; total solution coverage ≈ 0.85. • Low recommendation: Low Functionality was present in all low-recommendation paths; younger users appeared in low-recommendation combinations. Total solution consistency ≈ 0.87; total solution coverage ≈ 0.70. Demographic influences: - Although age and gender were not significant in regression, fsQCA revealed they participate in sufficient configurations: women and older users feature in high-recommendation paths; younger users appear in low-recommendation paths. Managerial implication from findings: - Enhancing engagement and ensuring accurate, credible information can drive recommendations, particularly among women; preventing usability/navigation issues (Functionality) is critical to avoid negative recommendations.

The findings indicate that perceived app quality significantly shapes users’ recommendation intentions, but the impact depends on how variables combine. Linear regression identified Aesthetics and Information as strong individual predictors, suggesting that visual appeal and the credibility/relevance of content bolster recommendation intentions. However, regression did not capture the role of Engagement or demographic factors. fsQCA revealed that high recommendation emerges through specific configurations, notably the combination of female gender and strong Engagement, with Information frequently contributing. This underscores that user segments may respond differently to app attributes: for some, engaging features drive advocacy when paired with demographic tendencies. Conversely, low Functionality consistently appears in all low-recommendation paths, highlighting that usability and smooth navigation are minimum requirements; deficits here undermine recommendation regardless of other strengths. The results address the research questions by showing that MARS dimensions—particularly Aesthetics, Information, Engagement (configurationally)—and demographics (age, gender) in combination influence recommendation (RQ1). Age and gender do not have uniform direct effects but participate in sufficient configurations that explain both high and low recommendations (RQ2). Practically, managers should strive for engaging, informative, and user-friendly apps, with tailored strategies recognizing demographic differences; methodologically, combining symmetric (regression) and asymmetric (fsQCA) approaches yields a richer understanding of user perceptions.

- Women who frequently use and find the app engaging provide the most positive recommendations. - App quality must deliver perceived value and meet user expectations (notably through credible information and appealing design) to foster loyalty and recommendations. - Age is a significant configurational factor: older users feature in high-recommendation combinations, while younger users appear in low-recommendation paths when functionality is weak. - Using both linear models and fsQCA provides complementary insights; fsQCA helps uncover multiple pathways leading to the same outcome, aiding more precise managerial actions. Future research directions include integrating more variables (e.g., satisfaction, service loyalty), exploring different service contexts (public vs. private, cross-country), comparing multiple apps, and employing more complex algorithms that account for user pathologies or preferences.

- Personalization limits: Fitness apps may not capture individual-specific needs, objectives, or user pathologies, potentially leading to inappropriate recommendations. - Sample scope: Modest sample size (N = 210) from users of a single private fitness centre chain within one country (Spain), limiting generalizability. - Variable scope: Focused on app quality dimensions and recommendation; did not include related constructs such as satisfaction or loyalty. - Measurement context: Cross-sectional perceptions; results may vary across different apps or service models. Future studies should use larger, more representative samples, compare public vs. private centres, include additional variables (e.g., satisfaction, loyalty), and test multiple apps and contexts.