Today algorithms are ubiquitous. They determine what advertising we see online, whether banks will lend to us or whether we will be picked for an interview by an employer. In the realm of audience measurement, algorithms are used to help address some of the limitations of pure survey-based research. Ultimately, they help marketers to more accurately assess the reach of their advertising messages.
All marketers want to know how best to reach their most promising prospects. But participants in surveys cannot be expected to tell us everything we want to know with perfect recall. As a result, we have developed a number of statistical techniques to help us get more value from our surveys. These range from simple survey weighting to ensure all parts of the population are properly represented, reach and frequency modelling to project audiences into the future and fusion and ascription to join disparate survey data together into an integrated view.
This paper explores some of those techniques.