Techniques are described in which a service operates to identify consumers corresponding to visitor interactions with resources available from a service provider. Features are selected to use for matching of clickstream data collected for unknown visitors to profiles established for known visitor IDs. The features are selected based on analysis that accounts for consistency, completeness, and uniqueness of features among a corpus of profiles. Then, relevance scores are computed over the selected features using an information retrieval model in which clickstreams are treated as queries and profiles are treated as documents. Unknown visitors are matched to corresponding profiles using the relevance scores. Access to the digital media content is then controlled in accordance with the matching based on relevance scores, such as by serving individualized marketing offers and content to consumers that is targeted to characteristics of the consumers indicated by respective profiles.