At Deductive, we focus on applying data science to diverse problems and opportunities in media markets. We work with networks, agencies, ad-tech providers, and ad buyers to help them insource datasets and put them to work. We are currently active in projects to activate ACR data, STB data, and Nielsen AMRLD data for various agencies, brands, and networks. We don’t claim to be the only people who have made this work. We certainly don’t have all of the answers. We have, however, now done this many times. We are sharing this white paper entitled Best Practices, so others in the industry may benefit from our experience and, hopefully, not repeat our mistakes.
We have distilled our knowledge of how and when to combine datasets in this free white paper that you can download here. We explore the three types of datasets, i.e., Panel Data, typified by Nielsen, set-top-box (STB) data, and data from Smart TVs. While some of the challenges and problems we find solutions for can be achieved solely using one type of dataset, others need a combination of two or three of the dataset types. We examine whether to combine datasets into a single panel or keep them separate and how to combine them by making them consistent. We explore the truncation or edit rules used to achieve consistency. We also delve into deduplication for when viewers have both an STB and a Smart TV. We then look at the highly significant issue of how to ensure combined datasets are nationally representative, focusing on geographic and demographic effects, and the impact privacy and opt-out regulations may have on constructing this nationally representative picture. We go on to probe program rating viewership using combined datasets, including live network, live program and time-shifted ratings, the problems arising from working with changing populations (changes in the way populations access TV, e.g., cord-cutting, etc.), household and individual viewing (not necessarily the same thing), and modeling household level viewing.
The White Paper concludes by looking at advertising ratings using combined datasets, including using them as run logs, native smart TV ad detection, measuring addressable, local and MVPD inserts, and exploring best practices when it comes to the critically important area of ad measurement, and how Smart TV native ad detection is the best way to measure advertising because it can cover not only national and local but also MVPD and addressable too.