Three ways digital advertising must evolve as we adopt AI
Anna Forbes, General Manager UK at The Trade Desk
It’s true that AI is proving itself to be a crucial tool, but there are a few things that need to change if advertisers are going to set themselves up for success
Despite constant media attention about the “rise of the robots” the advertising ecosystem has finally realised that AI is, in fact, a force for good. And rightly so.
Let’s think about the programmatic world we live in: to buy the best possible ad out there, we need to evaluate nearly 9 million impressions per second. And analysing each of these ads, deciding which is best, and then purchasing the top choice – all in less than second – is beyond human capability.
Not for a machine. Over the past few years, we’ve developed AI technology that’s so sophisticated, it’s able to flick through every advert out there to find the optimal one. And by working in harmony with this technology, advertisers are enhancing their lives. No longer do they need to waste time on boring processes and number-crunching – they can focus on what humans are best at: being strategic and creative.
But let’s pause for a moment. It’s true that AI is proving itself to be a crucial lifeline, but we need to make sure the synergy between man and machine is absolutely on point and that we’re not teaching tech any nasty tricks. We’ve had our eye out, and noticed a few things that need to change so that advertisers set themselves up for success:
1. Quality isn’t the only measurement
As an industry, we tend to gravitate towards quality as the ultimate measurement for campaign success. Viewability is the KPI buzzword, and the more video dominates the scene, the more completion rates become the ideal.
The risk is that brands – and agencies – sometimes use quality as the only marker of success, rather than a combination of different goals. And if you’re employing AI, this can have fairly detrimental consequences.
Let’s take interstitial ads – ads that appear whilst a new page is loading – as an example: they’re highly viewable because there’s no way of clicking away from the ad. If the super intelligent tech knows it’s being rewarded for viewability, it’ll automatically push all ad spend into these ad types. While it might win top marks for viewability, it could flop on what the campaign was really after.
Combining goals will mean you hit the jackpot every time. If you take, say, viewability teamed with on-target CPM, you’ll know you’re getting through to the right audience and serving them viewable content.
2. Last-touch attribution is a thing of the past
Today, last-touch attribution is the standard, but it’s time we changed that. The super intelligent machine will quickly understand that last-touch provides a significant loophole for corner-cutting. It’ll know that it can serve an ad to a consumer in a fraction of a second before they buy a product online, perfectly executing last-touch attribution – and delivering “outstanding performance” for the advertiser.
Here’s an example of how this works. A consumer has been served a series of connected TV ads for a new pair of jeans, and slowly been won over. When he or she finally caves and goes to buy those jeans, the machine serves a final banner ad and it’s that ad that gets the reward.
But can it claim credit? Is it good marketing practice? It is great performance? The answer is no. So let’s get rid of last-touch attribution, and teach our machines to understand great advertising.
3. Sales data is the key to optimisation
In an attempt to move away from last-touch attribution, some advertisers use multi-touch tactics to find all the different credits along the consumer journey – which are then inputted into the machine. This is smart thinking, but because some businesses are operating within walled gardens, it’s not always possible to collect this data.
So, why not use real sales data? Our big supermarket chains, for example, analyse their sales data by postcode and product. If an advertiser had access to these insights, their ability to target effectively would soar. But if advertisers trained machines to optimise to this data, not only would their targeting skyrocket, they’d be able to move even further away from the nitty-gritty and start testing hypotheses. Which products work better in different areas? How far are our customers likely to travel to a shop? Sharing real-life data with our machine counterparts will pinpoint the perfect audience, and allow advertisers to get deeper into the strategy.
As AI establishes itself as the kingpin of the ad industry, advertisers will need to evolve their approach to digital media buying. Ultimately, AI is enhancing our advertising and allowing us to focus on what’s more interesting and important. But, we mustn’t pass our old, bad habits onto the machines - instead let’s teach them how to perform in the optimal way. Only when we achieve the perfect collaboration between man and machine will our advertising reach the highest level of efficiency, insight and relevance.