Where We Rate: The Impact of Urban Characteristics on Digital Reviews and Ratings

Abstract

In urban environments, eating and drinking out (EDO) is a widespread activity among residents and visitors, generating a wealth of digital footprints that reflect consumer experiences. These digital traces provide businesses with opportunities to enhance their services and guide entrepreneurs in selecting optimal locations for new establishments. This study investigates the relationship among urban spatial features, pedestrians and digital consumer interactions at EDO venues. It highlights the utility of integrating urban mobility and spatial data to model digital consumer behavior, offering potential urban planning and business strategies. By analyzing Melbourne's city center, we evaluate how factors, such as pedestrian count by sensors on the streets, residential density, the centralities and geometric properties of streets, and place-specific characteristics, influence consumer reviews and ratings on Google Maps. The study employs a random forest machine learning model to predict review volumes and ratings, categorized into high and low classes. The results indicate that pedestrian counts and residential density are key predictors for both metrics, while centrality measures improve the prediction of visitor scores but negatively impact review volume predictions. The geometric features of streets play varying roles across different prediction tasks. The model achieved a 65% F1-score for review volume classifications and a 62% for visitor score. These findings not only provide actionable understanding for urban planners and business stakeholders but also contribute to a deeper understanding of how spatial dynamics affect digital consumer behavior, paving the way for more sustainable urban development and data-driven decision-making.