Abstract: Aspects of the disclosure provide for methods, systems, and apparatuses, including computer storage media, for adaptively routing web traffic. A system can receive from a user computing device accessing a web resource, a content request, including data characterizing user interaction by the user computing device with the web resource before or after receiving the first content request. The system can identify a policy from a plurality of policies to execute in accordance with one or more objectives, wherein the plurality of policies are generated using data from a plurality of content requests and data characterizing user interaction with the web resource after serving a respective response to each of the plurality of content requests. The system can perform, in response to the content request, one or more actions of the ranked list of actions of the identified policy.