For all the fear of AI taking our jobs, there’s lot of optimism about the additional efficacy and effectiveness that AI can bring to ad campaigns. And I, too, hope for the renaissance of creativity that could be unleashed as we redirect our teams’ efforts towards creating ads, products, and experience that are more resonant, more memorable, and more persuasive. 

But one thing I don’t hear enough talk about is the need to be equally creative in the construction of the data that informs the AI. 

An unintentional consequence of digital ad measurement, like the ability to track clicks and cookies, is that many advertisers have spent the past decade developing tunnel vision on maximizing clicks on ads, at the expense of maximizing making something click in someone’s brain or in their heart. 

For AI to be truly successful in delivering the advertising apex we expect it to, we need to figure out how to give it a holistic view of human reaction – as well as action – to our ads. Put more data into the model. Get creative with the web of indicators we can weave. Design controlled tests to feed into the system in the absence of real time data. Be open to leveraging both expected and unexpected data sources as indicators of impact. 

This approach is necessary as we plan for the retirement of cookies and try to crack measuring the impact of traditional media, but it’s critical to solve before AI becomes the autonomous driver of our campaigns. 

So where do we start? 

First, the myth of a single source of truth needs to disappear. You were never allowed to turn in a college paper with only one source cited…and for good reason! The truth comes from finding patterns. Reliability comes from meta-analysis. We need to embrace the idea of triangulating on the truth across multiple sources. 

Second, we must invest in rethinking our measurement taxonomy and get creative about filling in the gaps. A potential approach includes:  

• Time-bound measurement: the known quantity is the moment in time when the ad was served 
• Location-bound measurement: the known quantity is where this ad served 
• Device/user-bound measurement: the known quantity is to which device or user the ad was served 
• Yes, we are going to have to deal with what’s in an olive garden and what’s outside an olive garden… or rather walled garden. I used voice-to-text and Siri interpreted “walled garden” as “olive garden.” Clearly, the tech has a way to go before people can stop being directors, supervisors, and editors! 
• And we need to invest in identifying the reliable ancillary metrics when none of the above are possible 

Third, we need to use this opportunity to invest in expanding the data graph on our ads to include creative variables. Does the ad feature people? celebrities? Is it lifestyle, product-focused, or demonstrating the product in use? Is it focused on product features, brand values, or price promotion? Does it aim to motivate through humor, fear, connection? 

As mature markets see population growth plateau, consumer spending power stagnate, and time spent with media stabilize, we may be reaching the point of diminishing returns on media efficiency, even with the seeming magic of AI to enhance our optimizations. For AI to be truly transformative at accelerating advertising effectiveness, we need to become unbound from today’s measurement models and get creative about revealing the full matrix of consumer reaction and action to the AI.