Structural drivers of sales in omnichannel retailing

In platform-mediated retail markets, store-level performance depends on how offline location fundamentals and online reputation signals jointly shape demand—yet whether their relative importance differs across fulfillment modes remains underexplored. We develop a channel-friction framework arguing that dine-in, delivery, and pick-up each embed distinct information frictions (differences in how consumers search, evaluate, and commit) that systematically shift the diagnostic weight of location versus review-based predictors. To test this framework, we implement a multimethod design. A preliminary survey confirms that consumers weight proximity, competitive conditions, and review cues differently by channel, providing demand-side evidence consistent with the friction logic. We then analyze weekly sales from 349 outlets of a nationwide Korean foodservice franchise (2021–2023), estimate rolling-window XGBoost models, and evaluate the incremental predictive contribution of each driver block across channels. Results support the framework: customer proximity is the dominant predictor across all channels; review signals provide the largest incremental gain for delivery—where decisions unfold in a low-inspection, digitally mediated environment—while contributing negligibly to dine-in, where experiential cues dominate. Within-chain proximity is more strongly associated with sales than cross-chain density, highlighting cannibalization risk. By integrating offline and online drivers within one channel-sensitive forecasting architecture, we demonstrate that predictive accuracy is highest when these inputs are modeled jointly and that the observed predictor heterogeneity can be traced to theoretically motivated friction differences across fulfillment modes.