Artificial intelligence (AI), in the form of machine learning, is being used across the industry to optimise programmatic digital campaigns against specific performance KPIs. Yet, while some elements of media buying are at the forefront of technology, much of it is still seemingly trapped in the dark ages.
Even online, we see people making manual inputs within trading desks across multiple platforms. As the landscape becomes increasingly complex, we’re simply seeing agency headcounts grow ever larger. Don’t get me started on traditional media, where people are still manning the phones and manually negotiating with TV channels and national newspapers. It’s slow, inefficient and mired in 20th-century thinking.
In the not-too-distant future, it would be nice to think that AI will help media planners analyse mid-campaign performance against sales to shift their plans accordingly. However, before that can happen, there needs to be a sea change in how we buy media. There is no reason why this programmatic approach shouldn’t encompass all media, not just digital, as long as there is more open access to real-world offline data.
In a perfect world, campaigns would become holistically biddable, where AI systems can bid programmatically for all media opportunities and not just the ‘programmatic digital’ ones. This approach would reward value and result in media inventory being priced for what the market will bear.
Roadmaps and dead ends
To put an auction-based, campaign-level AI platform in place, we first need to create a feedback loop that allows advertisers to attribute their campaigns against real-world action and not just clicks. Currently, quality data that links offline sales or footfall or other key performance indicators (KPIs) is, to put it diplomatically, flawed.
Real-time econometric modelling will help the AI systems of the future to connect purchase behaviour to advertising in a way that transcends the current boundaries. It will give offline a glimmer of the optimisation that programmatic allows digital to carry out so expertly when the end goal is a digital action.
We’re a long way from that, though. The landscape is fragmented and the best data is limited to specific aspects of programmatic buying. Offline data needs to be more accessible and we need to look at ways to build open application programming interfaces (APIs) and agree on a common set of industry standards with data owners.
To create a genuinely transparent and fair media market, we need to invest in AI platforms that allow buyers to book, manage and measure campaigns, and connect directly through open APIs with publishers, whether the inventory in question is a billboard in Covent Garden, a banner ad on HuffPost, an 8pm TV slot on ITV or a full page in next month’s Vogue.
Moving in the right direction
We’re seeing some baby steps – demand-side platforms (DSPs) using AI to identify data segments that are performing well and self-optimising, plus econometrics tools that can look at a single snapshot of the past and simulate the performance of a holistic media plan.
A ‘stock market’ model for buying media, however, needs richer data still and an open exchange for all media channels, where AI can respond to fluctuating changes in pricing, also means no need for people to do the ‘boring bit’ of actually buying insertions.
Why autopilot won’t be an option
The value that humans bring to media buying lies in planning and media strategy, and in defining the parameters to ensure brand safety and that its values remain intact: if the brand is all about quality rather than price, its AI needs to know to focus its real-time outdoor billboard ads on quality environments over cost in the short term, even if the latter seems the most effective tactic at a given time and place.
Context is hard for an algorithm and relationship building also lies beyond the capabilities of AI. Agencies will always need people to ensure client satisfaction. It’s not always rational and logical, and human-to-human interaction wins and retains more business than a thousand AI-driven trading desks, no matter how impressive the technology is.
We’ll have to do this as an industry, together, with our industry bodies as well. Ideally, AI will help get us to a world where an agency doesn’t have to call up a publisher to place one press insertion – and where the question clients ask isn’t ‘how much of your media plan is programmatic?’, but ‘how much of your media plan is AI-driven?’ That way, the models that are being designed and optimised to deliver the best end-goal in the online world can be exported to the offline.