Demand-response strategies for customized buses (CB) typically treat different passengers homogeneously, ignoring the heterogeneity between passengers in their choice preferences. In order to gain further insights into people’s choice behavior in CB and explore the influence of latent variables, this study combined the characteristics of CB operation, and latent variables reflecting convenience, comfort and reliability into a Mixed logit model to form a hybrid choice model. A survey was conducted on 30 CB routes in Dalian to study the operational characteristics of CB and quantitatively analyze the impact of various attributes on CB choice behavior using empirical data. The results show that CB is more attractive to middle-aged and young women with middle to high incomes. Reliability and traditional variables such as travel time have an equally important impact on choice, and the utility of average cognitive reliability is equivalent to the utility of reducing travel time by 6.15 min. The probability of choosing CB will increase by 1.09 times for every increase in the perceived reliability evaluation level. The hybrid choice model has a higher goodness of fit than the traditional logit model, and the prediction results are more robust. Travelers have heterogeneous preferences for travel time, waiting time, seat availability on regular buses, discounts on fares, and detour rates. To ensure a stable full-load rate, a discount of at least 12% on fares is needed to offset the perceived negative utility of approximately 90% of travelers, accounting for heterogeneity. Therefore, on the basis of providing multi-vehicle types and multimodal travel services, operators need to find a balance point between demand-responsive customized bus stops and route adjustments, and fare adjustments.