Artificial intelligence (AI) and machine learning are revolutionising the advertising landscape. From varying thousands of creative formats to optimising campaigns across screens and channels, AI is delivering marketing benefits that would have once been the stuff of fantasy.
That said, such powerful technology has aspects that marketers need to prepare for. As brands and businesses embrace this new approach towards marketing success, they need to consider a few key issues to ensure they maximise the remarkable opportunities AI presents.
Last-touch attribution, where the last ad seen receives credit for an online sale, is an unfortunate legacy of the early days of digital advertising. After decades of only theorising about the impact of a print, radio, or television campaign, measuring exactly how many users clicked on an online ad was a shiny new toy.
Regrettably, a lot of brands and agencies still treat it as such.
While they may also use other measurements, last-touch attribution remains the industry standard even though marketers now have the ability to be more precise about their goals.
Reinforcement learning is one area of AI that ‘rewards’ the machine for hitting certain goals. Last-touch attribution is one such example of this concept, but it can be a problematic application of it as the complex algorithms trained to master a specific goal don’t necessarily lead to good marketing.
For example, if an algorithm serves an ad to a consumer a moment before they buy the associated product, it would deliver outstanding performances for the advertiser. But, rather than driving true results, machine learning may simply have perfected the art of taking credit for good marketing without actually doing any marketing.
In practice this would be akin to someone who has been exposed to several months of branding ads for a new car model on their connected TV. But it’s the last online banner ad they saw before visiting the dealership’s website that gets all the credit – and we all know that’s not right.
Media quality isn’t everything
When it comes to judging the quality of advertising inventory, it’s crucial to ensure multiple measurements are in play. Agencies and marketers often become too fixated on specific goals, such as whether the ad was viewable or the consumer watched the entire ad. While important, these should never be the only goals. We’ve all seen examples of ads that were viewable but featured content ultimately detrimental to a campaign’s objective.
AI is very good at finding the best options for whatever parameters are provided, but if you only give it one subset of a goal – for example, maximising viewability – you may not get the best business outcome.
For example, in-app interstitial ads tick the box for viewability, but without the right guidelines an algorithm acting alone could potentially miss the strategic mark by pushing the majority of campaign spend to in-app.
On the other hand, a campaign that uses both viewability and on-target CPM to measure performance is going to more reliably hit the right age and gender targets.
Clear goals are the key
The ability for marketers to optimise business results – rather than proxies for business results – by training AI is one of the most exciting opportunities this technology brings to the table.
Compared to the days when advertisers only really knew if a user clicked on an online ad and how many times it was seen, advertisers now have access to an incredible amount of information they can include in their campaign.
Some clients use multi-touch attribution to credit different touch points along the conversion journey before passing the scores back to train an algorithm. Others are tying their campaigns to sales data, including car sales measured by Oracle, retail visits counted by Factual or lead conversions registered inside Salesforce.
Machine learning is becoming a driving force in every marketing plan, but machines can only do so much. While the future of our industry is somewhat reliant on this technology, it’s never been more important for marketers to have precise goals. Humans should drive the strategy of the campaign.
Change is happening but, like all revolutions, it will take time for everyone to fully appreciate its breadth.
It will be a slow burn only realised when the generation of new marketers – those who know the challenges of last-touch attribution and single-focus measurements – rise up through the ranks and land in senior positions in agencies and brands.
In the meantime, our job as an industry is to educate clients. We need to be getting on the phone and out of our offices to show brands there are better ways to measure campaign success.
Part of that discussion should be reinforcing that it’s not all about AI. Technology isn’t at a level where it can interpret real-world factors, so artificial intelligence won’t achieve optimal results without human intelligence behind it. Simply put, an engine is nothing without its driver.
The Trade Desk AUNZ GM Mitch Waters