Indie agencies focus on optimizing media and creative using AI to pick up more clients
By Michael Bürgi • October 22, 2025 •
Ivy Liu
Among the many ways generative AI and its machine learning ancestors are helping media agencies do their jobs more efficiently, campaign optimization seems to be catching on in a bigger way. Essentially, more and more media agencies have boosted and sped up their AI abilities to apply marketing mix modeling to media campaigns — and now it’s spilling over into creative as well.
The latest example comes from MissionOne Media, the newish media agency arm of Barkley OKRP, which is launching M1 Refinery, a data-driven insights analyzer of creative performance. M1 Refinery, according to its creators, aspires to answer the question: why does one ad perform better than another? While A/B testing and standard platform reports can show which ads perform best, they rarely explain why. M1 Refinery aims to do just that, both from a creative and media POV.
Using predominantly Google Gemini meshed with its own developed tech (the agency describes itself as a big Google partner), M1 Refinery analyzes up to 200 creative elements – such as product type, color, emotion, props, location, language and video length – across platforms including Google, Meta, TikTok, and YouTube. It then maps the creative elements to real campaign KPIs, to determine what actually drives results.
“We could actually pull out the elements of creative that are working for a brand across platforms and identify how they work across [any number of] KPIs, starting with lower funnel but moving towards upper funnel —what are the elements that actually work,” explained Sean Corcoran, president of MissionOne Media. “That gave an insight to us and to our clients in a way that’s different from any of the tools out there that are built for particular platforms.”
Early testing of the tool showed a 22% lift in ROAS for an unidentified beauty client’s campaigns across Meta, and 20% higher CTR on average after a single round of applying M1 Refinery to the campaign.
“We can work with creative here to bring them to the front lines, which we love, because I feel like they’ve always been kind of behind the front lines in the last decade, while we dealt in media and pipes and data,” said Corcoran. “This brings them to the table to see what’s working — and not working — for their work, and brings them into the mindset of optimization. It also brings us that larger, universal view of what’s working for a brand.”
Matt Zeiger, svp of technology at MissionOne Media, who came aboard with Barkley OKRP’s acquisition of Adlucent performance agency last year, noted that some of what the tech can uncover in creative insights is, well, unexpected. And he offered an example of that.
“One surprising insight is, we analyzed all these video ads that were in this [unidentified] account for years, and we were able to see, that it ends up that house plants in the background were driving a more of a positive predictor of a higher ROAS,” said Zeiger. “Who would have thought, right? It’s that type of surprising [insight] … So then it was like, Well, hell, let’s make sure we put more house plants in our video ads.”
Swapnil Patel, co-president at new media agency Attention Arc (part of Cheil Agency Network) who spoke at Digiday’s Media Buying Summit last week along with Sylvain Tron, managing director at sibling company Cylndr Studios, said the goal is to get these exact kinds of answers for clients.
“There’s a conversation we can have with Sylvan and team and say if we wanted to test this in this category, or templatize this, where we know these three variables matter, but we can’t get enough variation, is there a way to do that?” said Patel. “And that’s how a lot of our last six or seven months of conversations have been sparked — it’s what can we do? And if we learn from that, then we can decide to make more.”
“A lot of our conversations are media-led, so when the team sees performance variance by message or by targeting tactic, as a media person you’re often optimizing that for media performance,” continued Patel. “What we’re trying to do is take some of those learnings and bring them back to the strategy and creative and production teams so that they can say, ‘OK we’re seeing that this audience is resonating better with this message. Let’s ideate off of that again.’ Then you have to look at what’s possible production wise.”
Along a similar line, Ars X Machina rolled out a new AI-powered MMM model it dubbed Agile Mix Modeling — signaling a through line across independent media agencies to shore up their analytical skills across media and creative.
The timing is interesting, not only because it coincides with use of AI, but also because independent agencies expect to enjoy a windfall of medium-sized clients to drop out of the holding companies as the latter get bigger and bigger. Time will tell if this bears out.
