This paper examines how learning and competition shape brokerage commission fees in a plain-vanilla Chinese setting. We document three facts: substantial price dispersion across investors, higher fees for naïve investors, and widening dispersion over time. To explain these patterns, we develop a dynamic model in which brokers compete for naïve and sophisticated clients while learning types through repeated interactions. The model shows that an incumbent’s information advantage can sustain persistent market power—even over sophisticated investors—and that additional broker entry can backfire by raising markups on naïve investors and widening dispersion. Exploiting branch-level variation in rival presence, we provide causal evidence consistent with these predictions. Quantitatively, although the average number of brokers per city more than doubled over the sample, intensified competition slightly offset—rather than accelerated—the secular decline in fees.