Data Fusion
ObjectiveThere are many surveys that study the media consumption of people, for instance, TV panels which study TV viewing and print panels which study newspaper and magazine consumption. However, evaluating multimedia advertising campaigns is not as easy — one would require a single survey which studies consumption of TV, radio, print etc for every respondent and such studies are not available, due to practical and methodological restrictions. Pointlogic is working on a process called data fusion, in which several surveys are amalgamated into one single, artificial survey where the multimedia behaviour of the ‘respondents’ of the artificial survey is known.
Approach
Data fusion requires a matching process, with respondents from one survey matched as well as possible to respondents from others. It also, therefore, requires a ‘scoring procedure’ to define which respondents are alike enough to be matched, as well as an assignment algorithm i.e. a very specific discrete optimization problem. This project involved the development of a better assignment algorithm.




