Introduction

Identity fragmentation poses an increasing threat to the promise of big data in marketing analytics. Although event level records of advertiser and consumer actions have opened the door to a plethora of individual and household level modeling techniques, these fundamentally rely on the ability to integrate data recorded across different domains. Key players across search, display, social networks and browser software are increasingly working to disrupt the ability to record events in a common identity space, leading to an almost total reliance on ID graphs to bring big data into usable analytical structures. For example, a typical advertising campaign may involve TV impressions recorded against an IP address, display impressions recorded against a cookie labelling system, and outcomes such as sales recorded on a client's website or CRM system. The purpose of the ID graph is to draw together and integrate observations from all of these separate ID spaces such that the inputs and outputs associated with the correct individual or household can be associated.