This study aims to investigate the key drivers of restaurant choice by examining how consumer preferences vary across different situational contexts, such as meal type, consumption format and day of the week, and thus explore the interplay between hedonic and utilitarian motivations in everyday dining scenarios.
A sequential exploratory mixed-methods design was adopted with a Brazilian urban sample. First, 20 in-depth interviews were conducted to identify key decision-making dimensions and meaning cores. Second, 384 context-specific scenarios were generated using large language models (LLMs) to simulate consumer preferences across diverse sociodemographic, behavioural profiles and situational contexts. Third, a vignette-based survey of 247 Brazilian consumers was used to test the relative importance of attributes across eight controlled scenarios, analysed using general linear models.
Across all methods, cleanliness, taste and value emerged as the most important attributes. Contextual variation influenced attribute salience: utilitarian drivers (e.g. healthiness, speed, and cost) were more important during weekday lunches and delivery orders, while hedonic attributes (e.g. ambience, empathy, and culinary presentation) were emphasised during weekend dinners and dine-in occasions. Triangulation across methods revealed strong convergence on core themes, while also highlighting method-specific nuances.
This study integrates human and machine-generated insights into consumer behaviour. It offers a novel application of LLMs to simulate decision-making patterns and demonstrates the value of triangulated methods in capturing context-sensitive preferences. The findings offer actionable insights for restaurant managers aiming to tailor services to specific dining occasions. The use of LLM-simulated personas introduced innovative perspectives, although limitations in cultural nuance and interpretability should be noted.
