The future of targeting in a privacy compliant age
Chris Turner, the Digital Investment Director at MediaCom North, on why it’s time to look to the future and stop longing for the ‘good old days’ of targeting.
Once upon a time in a galaxy far, far away life was relatively simple. Targeting was tied by and large to content/context or 3rd party cookies, it was most likely to be across two devices (work and home) for behavioural understanding, and ad blocking didn’t really exist at scale.
Life was simple, how we long for those days… actually, we don’t - and for good reason.
Data for targeting was about as opaque as a block of concrete and the digital experience was about as rich as dried toast. From an application and measurement perspective, this was usually falsely tied to direct response outcomes via cookies. These cookies were dropped in the right place in a converting path, becoming the proxy for success - which with it came sub-optimal practices and negative experiences.
Whilst the ecosystem has become far richer across the past 20 years, it has also become markedly more complex to deliver seamless cross channel/content targeting. We saw an increase in the popularity of single entity platforms or forms of consumption that did not rely on or use 3rd party cookies at all. Instead, these used different data sets and infrastructures which did not ‘talk to one another’.
Therefore, as we move firmly into the era of privacy, the challenges created by the deprecation of 3rd party cookies aren’t really, when one reflects, new challenges at all - they are simply being brought more visibly to the fore. The concept of total audience utopia - “we know who they are, where they are, what they like, when we should talk to them, what we should say, across all devices and touchpoints” – hasn’t really existed at total scale.
The era of privacy simply forces us to re-evaluate, innovate and rearchitect how we can target users.
Outside of the known walled gardens, looking to the future at an apex level, there will be three primary routes to privacy compliant deliver, not all of which are new concepts.
Audience built from aggregated data
These are essentially de-identified data sets which are privacy compliant and can be actuals in totality or modelled.
In the case for the former, at a very salient level think of scalable mobile device data which is delivered at a postcode level into cohorts/clusters. This can then in turn be used to provide insight around the digital and physical behaviour of audiences, but rolled up, as opposed to one-to-one - informing targeting and media deployment. Skyrise Intelligence is an example of an operator in this space.
In terms of modelling, this already happens via 1st party seed data, commonly known as look-alike modelling. Machine learning will help further enhance modelling, two examples of which include:
- Agent based modelling creates simulations of how populations (or audiences) might behave based on different scenarios, and what will happen when variables in the ecosystem change. This uses virtual humans, hence the ‘agents’, and allows the creation of synthetic audience segments without the collection of personal data.
- Federated learning uses the individual devices to host a machine learning algorithm, so the data is kept local, at a device level, and therefore in principle, personal data isn’t shared back to a central server for service analysis. These are some of the principles behind the Google Privacy Sandbox
Audience defined by content & context
The industry has regularly fluctuated between broad polarised statements such as ‘audience is key’ and ‘context is king’, the logical reality being it’s a bit of both and they are not mutually exclusive points.
Contextual delivery comes in various guises - from literally selecting content environments and semantic keywords to more advanced solutions. One such solution is illuma, which crosses over into AI and modelling by essentially understanding the pages/content which generate positive engagements from your audiences, while looking for similar context or content. In all instances, these do not rely on 3rd party cookies.
Audiences build from consent
The value, role and use of 1st party data from its given source, with clear consent, has never been more heightened than is currently the case. There are three primary sources:
- Consumer Data Platforms (CPD) tie client 1st party data and create a deterministic ID to collect web analytics data. These are primarily focused on known customers. For clients/advertisers, scaling 1st party data through increasing consumer engagement with the clients/advertisers’ assets will be key to increase the number of known customers and what you know about them.
- Increase in paywalls and/or login walls for web services and content which is in essence a login, paid or not, that will create a declared identifier as a data asset.
- Publisher media and data alliances – media co-operatives and data alliances. In the UK you can think of the likes of Project Ozone, which brings together the 1st party data of some of the leading publishers/publishing groups in the UK into a shared ID and creates singular behavioural segments across them. This allows you to target a consumer using the same data set across publishers – so in principle high calibre, consented data coupled with the scale of all of those publishers, bringing together audience and context.
In the case of the latter, the likes of Infosum and the ‘data clean room’ concept will likely further democratise this practice, facilitating privacy compliant unification of audiences from 1st party assets.
There is a lot to be positive about moving forwards. Whilst making the task more challenging, protecting consumer privacy will ultimately raise the bar, removing bad actors, improving quality, and accelerating development in areas such as machine learning and AI.
Chris Turner is Digital Investment Director at MediaCom North.