What’s the last ad you remember?
Most people would be hard-pressed to answer that question.
Genero, a creative and content production platform, is working with creative testing startup Memorable AI to figure out what makes some ads stickier than others.
The duo has partnered on pitches to prospective clients, including L’Oréal, using AI to refine storyboards and concepts. Memorable AI brings scale and speed to the creative process, said Matt Perry, SVP and managing director at Genero.
The goal, said Memorable AI CEO Sebastian Acevedo, is to increase the return on ad spend by making a client’s creative more effective.
And the AI can deliver surprising insights along the way.
For instance, Perry said, an image with an overly busy background detracts from brand recognition, because there’s too much going on for the human eye to understand in a few seconds’ time. This drags down an image’s memorability, saliency and branding scores.
But AI can pick up on the busyness in a frame quickly and recommend ways to simplify the image, such as by going from two characters to one or using clearer messaging.
“Humans can do that, to a certain extent, but it would take a long time, and you’d never be really sure,” Perry said. “And how could you scale that up to thousands and thousands of individual frames across thousands and thousands of assets?”
Once a client accepts its pitch, Genero plans to work with Memorable AI on the project throughout the production process.
They’ll test creative assets along the way to determine what’s working in a given frame, sequence or section. By pairing those tests with heat mapping and analytics, they’ll produce final assets that suit the client’s needs and (ideally) outperform their current assets.
Remember me
So, what makes creative stick? Sometimes, it’s the simple things.
Round shapes, for example, are more memorable than straight lines.
This observation was one of many that Memorable AI derived from the cognitive experiments it runs, Acevedo said.
Another “very obvious, almost evolutionary” observation is that a human face is more memorable than the shape of a mountain, he said.
Memorable AI’s experiments focused on identifying what types of visuals attract more attention in the crucial window between zero and two seconds.
Those first couple of seconds are “shaped by unconscious patterns,” Acevedo said, as people tend to look in the same direction and remember images in the same order.
Memorable AI uses these and other related findings to predict the memorability of an asset and fine-tune its models, including tapping into performance data from ad management tools – such as from Meta, Google and Amazon DSP – and historical data from clients.
Its deep learning models analyze an ad’s background, featured objects, people and actions taking place, how noticeable the brand’s product or logo and the emotions present in the ad. Memorable AI even takes into account the cognitive load an ad requires, meaning how hard it is for the human brain to process the different messages and visual content in the ad.
Based on the results, Memorable AI can tell brands which creative elements are working best.
More memorable assets reduce the cost of brand recall lift and cost per click, according to Acevedo, and also greatly increase video view rates.
Checking all three boxes
The old saw in advertising has been that if you want “fast, cheap and good,” well, good luck. In the past, you could only have two out of three. But marketers today expect fast, cost-effective and good as the standard.
“If you’re not able to do that every single time, at scale, with budgets that are nothing like what they used to be, then you’re in real trouble,” Perry said.
For Genero, which is looking to expand its client roster, one of the main benefits of working with Memorable AI is the “agility” its technology enables. Genero can get a speed read on what’s working – and what’s not working – in its creative on “a relatively modest budget,” Perry said.
“How can you be better than 10 million data points? You can’t,” he said. “It’s not possible that one person has got a better idea than 10 million data points, so let’s respect that and use it and apply it.”
