We describe a model for predicting subjective user satisfaction based on objective ride conditions. The model takes ride features (e.g., duration, crowding, waiting time) as inputs and estimates users' single-ride satisfaction. This model enables transit authorities to predict the effect of service changes on user satisfaction. This in turn makes it possible to compare multiple candidate system configurations to determine which would yield the highest level of satisfaction. Using a sample of mass rapid transit users (n=641), the model is first trained then validated against unseen data. Prediction uncertainty is accounted for using a residual term. We also provide an example of how this model can be used to derive predictions, using artificial data. Highlights We derive a predictive model of single-ride satisfaction from a diverse sample of transit users (n=641). Transit authorities can use this model to predict the impact of changes to service on rider satisfaction. Validation analysis is provided to confirm the model's applicability to new cases. PREDICTIVE MODEL OF TRANSIT SATISFACTION 3

Not all transit users have the same preferences when making route decisions. Understanding the factors driving this heterogeneity enables better tailoring of policies, interventions, and messaging. However, existing methods for assessing these factors require extensive data collection. Here we present an alternative approach - an easily-administered single item measure of overall preference for speed versus comfort. Scores on the self-report item predict decisions in a choice task and account for a proportion of the differences in model parameters between people (n=298). This single item can easily be included on existing travel surveys, and provides an efficient method to both anticipate the choices of users and gain more general insight into their preferences.