Read the original article on NewDigitalAge here.
Traditional ‘lookalike’ audience modelling has always had something of a mixed reputation with brand advertisers. While some people regard it as a great way to add scale to their digital campaigns, others dismiss it as machine-learning-based extrapolation of limited value, often used simply to bulk up a campaign’s numbers. Based on the experience of brand advertisers to date, you could probably make a case for either argument.
However, like so much of the contemporary adtech marketplace, lookalike modelling as we know it will soon become a thing of the past, thanks to the final demise of third-party cookies. Traditional lookalike modelling methods work by taking relatively small ‘seed’ audiences (for example, male Guardian readers) and using the benchmark characteristics of these seeds to build much larger audiences for advertisers to target. Unfortunately, in practice, this effectively means matching one set of cookies with many others that resemble it, so the practice is living on borrowed time.
Without cookie-to-cookie mapping to rely on, advertisers will have fewer and fewer people to target and match with. As a result, if I were CEO of a Data Management Platform (DMP) right now, I’d be worried, particularly if most of my business is effectively based on selling ‘bags’ of cookies, with a few more cookies thrown in for good measure. Many are now embracing one or more universal ID strategies, but it’s unlikely these will provide anywhere near the scale or completeness of the third-party cookie they’re looking to replace, at least in the near term.
At Blis, we believe that the changes taking place in adtech are not only good for the consumer from a data privacy perspective, but will also ultimately be beneficial for the industry, both in terms of its performance and its reputation. Take for example, our own proposed solution to the audience mapping problem: affinity modelling to create dynamic audiences.
Introducing affinity modelling
Blis is tackling this challenge head on by leveraging its prowess in the location-based advertising space. Our new suite of privacy-first products includes our Dynamic Audience Targeting tool to reach personalised audiences at scale, all without reliance on personal data. It combines location data with hundreds of aggregated and anonymised behavioural and lifestyle signals, building a more complete picture of a brand’s customers, to target them in a privacy-first way.
Take for example, Ikea customers, of which there are millions all over the world. For each store, we could look at the postcode regions of existing customers, then learn the average income in that area; we could then layer that with shopping data from a global credit card provider, establish the socio economic class of the average Ikea customer, their preferred media channels and so on. By figuring out the shared characteristic or traits of existing shoppers, we can then use that anonymised, privacy compliant profile to match with others online who share the same mix of characteristics.
While this kind of affinity modelling shares some similarities with lookalike modelling from a machine-learning perspective, it represents progress on several levels. The new affinity approach involves extrapolating audiences from the anonymised characteristics that define them, then finding those characteristics dynamically, and refreshing the data in real-time. This is only possible when you have a top-to-bottom, integrated data and bidding platform to seamlessly do the matching. This integration allows very complex, multi-dimensional characteristics that closely match the original audience – not just simple IDs or lists that can be exported into a DMP.
By rejecting the cookie-based tracking and audience mapping techniques of yesterday in favour of privacy-first Dynamic Audience Targeting, the industry can once again start to leverage context and creativity to understand, locate and reach its audience, rather than simply rely on cookies to stalk consumers online. At Blis, we believe that the privacy-first future and affinity audience modelling could end up producing better advertising for everyone and we’ll be playing our part to make it happen.