In experience-based risky choice, people can learn about the options’ payoff distributions by drawing samples from them (ecological search). When the distributions’ key properties are not directly represented in the sequence of samples, the cognitive system needs to further process the samples to form such representations (cognitive search). The ecological search pattern (e.g., switching more or less frequently between options) imposes an initial organization on the samples, which can facilitate or complicate cognitive search by altering the number of processing steps required for a representation (transformation distance). Adapting one’s ecological search to reduce the transformation distance might be most important when more complex choice ecologies impose higher processing costs. In a preregistered laboratory experiment (N_Particiants=183; N_Trials=7,320), we tested whether participants adapted their ecological search to facilitate cognitive search in light of a given decision goal and choice ecology. Participants should choose either the option with better short-term or long-term rewards in choice problems of varying complexity (number of outcomes per option). The results show that ecological search shapes choice accuracy as predicted from differences in transformation distance: Participants who switched more were credibly better at identifying the short-term winner and worse at identifying the long-term winner. Moreover, participants’ ability to correctly identify the short-term winner was more strongly associated with switching in complex problems. However, participants tended to switch infrequently overall, with little difference between decision goals and complexities, indicating limited adaptivity in ecological search.