Predicting the intent of sponsored search users: An exploratory user session-level analysis

Over time, an online user searching for information about an idea or product may enter multiple search engine queries, thus creating a keyword search pattern from which the user's intent may be inferred. Such inferences could lead a merchant to alter the messages or provide offers to push the user toward a purchase decision once the user reaches the advertiser's website. Our research seeks to establish the relationship between these patterns as they occur during a user's search session and the user's purchase behavior. To test our hypotheses, we examine a unique dataset from a large Asian travel agency that includes over two million unique search engine queries and clicks as well as the same users' corresponding on-site behavior over a one-year period. We developed a typology for the coding of search queries used in determining the level of specificity and breadth as well as content type for each of the searches. Our analysis provides important findings regarding the relationship between search patterns and behavior. •Users' online search keywords give hints about users' purchase intention.•A keyword coding scheme was developed to analyze users' online behaviors.•Keywords analyzed by sessions and across-sessions•Users' purchase intention predicted using statistical and machine learning techniques