Review Topics Driving Star Ratings in Food Delivery Apps: A Cross-Country Comparison of Baemin (Korea) and Grab (Indonesia)

Abstract

This study compares the latent themes of online consumer reviews and their effects on customer satisfaction, operationalized as star ratings, between leading mobile food-delivery applications (FDAs) in two Asian markets: Baemin in the Republic of Korea and three Indonesian super-apps that all bundle ride-hailing with food delivery — Grab, Gojek, and Maxim. Because the Indonesian super-apps pool reviews across functionally distinct services, we apply a lexical-filter procedure that retains only food-delivery-related reviews (containing tokens such as makanan, pesan, grabfood, gofood; excluding ride-hailing tokens such as taksi, grabcar, gocar). Five hundred reviews per app were collected from the Apple App Store on April 29, 2026; the food-delivery share of the review channel was 38.0% for Grab, 13.0% for Gojek, and 7.0% for Maxim, indicating that the ride-hailing-anchored super-apps (Gojek, Maxim) carry far less food-delivery signal than the food-delivery-leading Grab. After filtering, the analytic samples were N = 476 for Baemin and N = 290 for the combined Indonesian food-delivery corpus. Latent Dirichlet Allocation (LDA) extracted five topics per market; star ratings were regressed on the document-topic proportions with HC3 robust standard errors, and the Indonesian model included app fixed effects to absorb baseline differences across the three platforms. Both models were statistically significant (Baemin: F(4,471) = 7.53, p < .001, adj. R² = .058; Indonesia: F(6, 283) = 6.09, p < .001, adj. R² = .094). For Baemin, delivery-time delay (B = -1.01, p = .002) and customer-service handling (B = -0.94, p = .005) emerged as significant negative drivers of star ratings. For combined Indonesia, the long wait / order cancellation topic drove ratings down significantly (B = -0.68, p = .025) once app fixed effects were controlled. Topic prevalences also differed across the three Indonesian apps in theoretically interesting ways: Maxim users voiced disproportionate concern with delivery fees and pricing (42.5% topic mass), while Grab and Gojek users emphasized waiting and cancellation. Theoretical, methodological, and managerial implications are discussed.