Data modelling: A powerful competitive advantage in the new data privacy landscape
Major shifts in the privacy landscape in the past couple of years are making it more difficult for brands to reach prospects and customers with personalised messaging.
Not only are regulators and legislators worldwide taking a harder line on data privacy, brands face a range of new restrictions on how they can use third-party cookies and mobile identifiers to deliver and track ads.
To close this new gap in the insight brands have into customers’ behaviour and inferred interests, platforms such as Google and Facebook are looking at new approaches for getting the perfect ad in front of the right customer at the most opportune moment.
One of the tactics they are looking at is not new — data modelling — which offers brands a powerful way of engaging with customers without the same reliance on cookies or mobile identifiers.
Before we delve deeper into data modelling, let’s review how digital advertising works in a world of third-party cookies. Most of us know that feeling of being stalked across the web by display or social media ads that seem eerily relevant to our tastes or interests — whether that’s the search for the perfect new office chair or finding a comfortable business hotel in Bloemfontein. These ads are served to us based on the websites we visit.
Each website tracks us via a small file called a cookie, which may contain data such as usernames, passwords, purchases and which parts of a website we visit — these are all examples of data collected in a first-party cookie. Third-party cookies are cookies that are shared across all domains that have the platforms’ tracking code. This allows the platforms to then track your behaviour as you move across these third-party domains, meaning the data is shared beyond the websites you are visiting.
The data is fed into advanced algorithms on platforms such as Facebook and Google, which use the data to determine which ad to show to whom and when. Using this data, brands can target the right potential customers with highly personalised ads based on their interests and behaviour — without needing to define an ideal target audience. However, this is coming to an end — and we need a new approach.
Knowing a customer better than their family?
In the new data privacy landscape, platforms are turning to data modelling, which uses partial data to make inferences about users’ potential willingness to respond to an ad. This isn’t a new idea, and it has been used for years in the offline world to powerful effect. An anecdote from Charles Duhigg’s 2012 book The Power of Habit relates how a retailer inferred from a teenager’s shopping habits that she was pregnant. Her unknowing father was furious when the retailer sent her coupons for baby products.
Google and Facebook are moving back to these types of techniques to close the gap between the data they have around conversions and the behavioural data they are increasingly unable to access from third-party cookies. But in a world where platforms have access to anonymised data about billions of users and the processing power to crunch massive data sets, they can apply far more sophisticated models than the ones companies used in the past.
In fact, these predictive models — which receive aggregated data to calculate probabilities using machine learning — can allow us to interact with customers in even more sophisticated ways than we did when we depended mostly on third-party cookies. Agencies should be upskilling themselves to help their clients use data modelling to address the move away from third-party cookies.
This entails building their own models and using application programming interfaces (API) to securely feed data into ad platforms such as Google and Facebook, which then better informs the machine learning algorithms. When brands are looking for digital agencies, deep technical skills such as data modelling and APIs should be high on their agenda.
An algorithm is only as good as the data it has available. Thus, data modelling can give brands a real edge, because they are not relying only on the same data and models on the platforms that everyone else can access. They can feed the algorithms with proprietary first-party information about their ideal customer that allows them to refine their targeting and messaging to improve return on investment.