We sat down with Chris Hogg, Chief Revenue Officer at Lotame, to talk about what’s new and what isn’t with AI, why a consistent identity strategy is essential for leveraging the technology, and what marketers should be putting in their data portfolio.
By R. Larsson, Advertising Week
Why are consistent, cross-platform signals important for successfully leveraging AI?
There is some naivete and wishful thinking about what AI can do with data. Machine learning models are fantastic at uncovering patterns in datasets, but you can’t just give one a pile of unsorted signals and hope that it turns lead into gold.
This is especially true when it comes to identity resolution. If you feed consumer data without an identity spine running through it into a machine learning model, it may be able to glean some connectivity between data points, but no model can create something from nothing.
What consistent, connected data reveals is not just the devices and domains a user has interacted with, but also the sequence in which these interactions took place. For example, a consumer seeing an ad on a travel app on their commute, researching the product on the web at work, then buying it in store via accrued loyalty points.
This temporal element of consumer data cannot be understood without identity resolution, and is so important for understanding how one touchpoint in a campaign affects the next. Such insights are fundamental to the effectiveness and utility of any predictive AI system that assists with audience intelligence or campaign strategy and execution.
In fact, I would argue that connected data is the very foundation from which the future of predictive, automated marketing can be built. How long have we been hearing about “a holistic view of the consumer”? AI applied to identity resolution gives marketers exactly that, with that view expanding from what a consumer has done to what they might do in the future.
What does “data connectivity” actually look like today?
For me data connectivity is about bringing together diverse consumer signals from multiple channels and platforms, while allowing for a certain amount of fluidity – as these signals will evolve as consumer behaviour changes. Despite Google’s shift away from the deprecation of third-party cookies, many brands had already made the decision not to rely on them any longer; turning instead to first-party data collection and activation. To enhance their owned data, brands are now leaning heavily on data enrichment and AI: tools that take basic customer data and uncover deeper insights or more effective targeting segments. It’s not yet a perfect practice, but it is a fast-evolving one.
Crucially, brands can’t do this work in isolation. They need to collaborate with partners who specialise in fresh, trusted data, identity resolution, and data collaboration technology. Echoing the IAB’s perspective that “the advertising ecosystem still requires multiple solutions to safely and effectively target consumers”; the same holds true for data connectivity.
Recent industry events have put AI front and centre, but few are talking about the groundwork that needs to happen underneath. Are marketers in danger of putting the cart before the horse?
For all the excitement around AI, it’s worth remembering the fundamentals haven’t changed, especially if we’re talking about machine learning which accounts for the majority of AI-powered solutions at this year’s Cannes Lions, for example.
Machine learning is already deeply integrated into data-driven audience intelligence and digital advertising, and has been for some time. You can go back to the turn of the millennium and find papers discussing how machine learning can be utilised for click prediction. Whenever you see the word “probabilistic”, machine learning has had a hand in it.
Machine learning is far more accessible now, and the interplay that natural language interfaces have with these models represents a massive leap in data accessibility. Garbage in, garbage out applies just as much today as it did two decades ago when machine learning was making its baby steps in advertising. If you don’t have robust, organised, useful data already, no AI model is going to fill the blanks for you.
If you do have your data ducks in a row, then AI can be a massive accelerator for your marketing operations. But you need to lead with a data strategy first before you put that data to work.
Why is it important for marketers to own their data strategy?
No one can own the entirety of their data strategy, that’s inherent in the need for connectivity. You have a few pieces of the puzzle, and must connect and collaborate with others to fill the full picture. Every publisher and brand, no matter how influential, only represents a thin slice of a consumer’s full engagement with media and commerce.
What all companies should do is own as much of the identity pipeline as they can. First of all, you need consumer data, whether that’s first, second, or third-party, then you need to integrate an interoperable ID solution that can connect this data to the wider ecosystem.
If you rely entirely on closed loop, walled garden platforms for consumer identity resolution, you are again limiting your view while also cutting off opportunities for mutually beneficial collaborations. There are ample opportunities for brand/publisher co-marketing and takeovers that can be unlocked through data collaboration, but not if you lack insights of your own to bring to the table.
Walled gardens, on the other hand, do not, by their nature, collaborate; it will always be a one-sided exchange.
How can marketers tap into behavioural, contextual, and declared data to power better outcomes?
I’m often talking to clients about the importance of having a varied portfolio of declared and demonstrated data, particularly when so many are still reeling from years of ever greater signal loss.
Each type of data uncovers a different aspect of consumer interaction. Behavioural data represents what a consumer has done, such as logging into a website, clicking on an advert, or researching a product.
Contextual data tells you about the media consumers are engaging with, including aggregated patterns of engagement over time, which reveals how media and its various characteristics affect advertising effectiveness and outcomes.
Declared data, or zero-party data – for example from surveys – is useful because it’s straight from the horse’s mouth. You have to take the responses on trust (people do sometimes exaggerate in surveys!), but it allows you to gather qualitative, brand-specific insights that can’t be found in ones and zeros.
Not every marketer will need every type of data, and the balance between them will vary on a case-by-case basis, but everyone should at least be familiar with their options and what each brings to the table.
