<p>On shared accommodation platforms, consumers usually rely on review information to reduce the uncertainty of their decisions. This study, combining econometric models and interpretable machine learning, systematically explores for the first time the correlational pattern between vertical rating dispersion (the inconsistency in ratings from different consumers for the same product) and horizontal rating dispersion (the differences in ratings for the same product across various dimensions) with the bookings of shared accommodations. Through an empirical analysis of five million reviews on Airbnb, the research hypotheses were verified. The study found that there is an inverted U-shaped relationship between vertical review dispersion and bookings. Moderate vertical dispersion can promote the growth of bookings, while excessive vertical dispersion will inhibit bookings. The higher the horizontal review dispersion, the fewer the bookings. The moderation effect analysis indicates that high review valence will exacerbate the negative impact of review dispersion, and strong review recency can alleviate the negative impact of review dispersion. The heterogeneity analysis shows that in the high-end consumption market, the negative impact of horizontal review dispersion will be intensified. Using XGBoost-TreeSHAP explainable machine learning technology, it was discovered that the proportion of 3-star reviews is significantly more important for predicting bookings than other ratings. For entire house rentals, higher ratings for accurate descriptions and cleanliness are associated with higher bookings; for room rentals, higher ratings for communication, location, and value are linked to higher bookings. The research conclusions provide a theoretical basis for shared accommodation platforms and hosts to optimize their operational strategies, and deepens the understanding of the consumer information processing in non-standardized service scenarios.</p>