Addressable TV - technology that advertisers can use to target TV ads at particular households or devices - is generating significant column inches within the TV industry. But GDPR is - at least partially - extinguishing some of the hype in Europe. It's the first thing most customers in Europe ask us about. [Addressable TV often relies on third-party data] (https://martechtoday.com/addressable-tv-state-cross-device-ad-buying-203737) to execute the buy, and third-party data is subject to some of the most stringent parts of the GDPR.
But we think that some of the examples we see in Germany - which already has strict, GDPR-like privacy regulation at a federal level - should negate some of those concerns. Germany has some of the most stringent privacy laws in the world, and many of the provisions found in GDPR already exist in German law. Yet we have seen big companies on the sell side, particularly ProSiebenSat.1, using probabilistic matching to target viewers in a way that adheres to the country's privacy laws
TV GDPR: Enter probabilistic matching
While more traditional ad campaigns do not require advanced data sets, the very nature of addressable TV means that sophisticated, high-quality datasets are needed to target the right people with the right ads. The critical challenge for businesses operating in the EU is to build and use these datasets without violating the new GDPR privacy laws.
Under GDPR, individuals in the EU have the right to view any data held about them regardless of the location of the company or the stored data and have the right to demand the rectification of their data. Individuals can also request the deletion of their data unless the company can provide a compelling reason not to do so, which it is unlikely to be able to do for data used in addressable TV. A company can argue that it needs to store the pseudonymized physical address and the IP address of its customers so they can receive products, make repairs, etc., but is not going to be able to argue that they also need such data for the datasets used in targeting ads.
This is where probabilistic matching comes in. Unlike deterministic matching, it does not use personally identifiable information but instead focusses on the likely characteristics of the households it wants to reach. One company we see doing this in Europe is ProSiebenSat.1, a free-to-air and pay-TV German broadcaster which has been particularly innovative in creating successful addressable TV campaigns using Smart TV data. PSS.1 models attributes based on IP location, combining this IP location with weather data, on technical features and attributes modelled from AGF, the German association providing the official TV audience ratings (i.e., the German Nielsen). These modelled attributes include viewing behaviours such as the demographics which typically watch particular genres or segments, e.g., women 18-44 tend to like romantic comedies, men of all ages tend to like soccer, etc. The technical features which PSS.1 use to help them create targeted ads include the platform the customer is using, the time of day a customer is watching TV and whether they are watching via DVB or DTH.
Retargeting is a technique which PSS.1 have developed with some carefully selected partners, who agree to share their data though they also use TV data gathered from their own services. GDPR demands that partners explicitly opt-in, but given the voluntary nature of the association this is easy to achieve. PSS.1 have been working on expanding their own range of digital assets because the company can then use the data which comes with these assets in the re-targeting process.
Our data science consulting team see PSS.1 as having the best model for GDPR compliance regarding addressable TV, and we expect to see the majority of TV companies operating in the EU region to follow their model in four specific ways:
- By targeting based on geo features such as where you live (not just the name of a street but factors such as average wealth in that area and whether it is raining, a weather factor likely to increase viewing).
- The technical features of the viewing device (so news programs are more likely to be viewed on a smartphone than the latest blockbuster film.)
- Building segments through modelling by using panel data and some small opt-in pools to match different viewing behaviours with different targeted ads.
Talk to our data science team if you want to learn more about probabilistic targeting and modelling