I produce a lot of content. A lot.
Case in point: I am scheduled to complete 19 interviews in the next two days at the Retail Technology Show in London.
So when I sit down with my team to decide what to push, what to feature, what to put energy behind, a significant part of that decision, like it or not, is gut feel. Experience, pattern recognition, instinct. It’s not nothing, but it’s also not science.
So when Armen Mkrtchyan, CEO and co-founder of Extuitive, started our latest Spotlight podcast conversation by reaching back to John Wanamaker, the 19th-century department store pioneer who famously said “half of my ads don’t work, I just don’t know which half,” I felt that in my bones. Because nearly 200 years later, most of us are still operating exactly the same way Wanamaker was.
We’re just doing it faster, at higher volume, and with a lot more money on the line.
The Problem Has Only Gotten Worse
Here’s what’s changed since Wanamaker’s era: scale and complexity. Today’s brands aren’t running a handful of print ads. Some of the partners Armen works with are launching hundreds of ads per week across Meta and TikTok. The platforms themselves actually encourage this behavior.
Create diversity. Launch more. Let us optimize, they say.
However, the platforms’ incentives are not perfectly aligned with yours. From their internal analysis at Extuitive, they’ve found that the very top-performing ads aren’t necessarily the ones the platforms promote with increased spend. It tends to be the second-quartile ads, i.e. the ones that perform well enough to keep you coming back and spending, but not so well that you’d ever question whether you needed the platform at all.
That’s not a conspiracy theory. It’s just incentive structure. And it means that if you’re relying entirely on platform optimization to figure out what’s working, you may be leaving real money on the table.
The AI Content Problem Inside the AI Content Problem
There’s another wrinkle here that Armen articulated better than anyone I’ve heard. AI-generated content, by its nature, tends toward the mean.
If you go to any large language model today and ask it to simulate how a particular consumer persona would react to an ad, it will give you an answer. But that answer will almost always reflect the average of what that type of person tends to do, not the true diversity of views that exists within any real population. You lose the outliers, the edge cases, the unexpected affinities that often turn out to be your most valuable audience segments.
This is why, as Armen explained, building predictive models on top of AI personas alone doesn’t get the job done. You need real data. And not just survey data bit behavioral data. Purchase receipts. What people actually buy, not just what they say they’ll buy.
How To Solves The Problem: The Fusion Model
Extuitive’s approach is built on two foundational pillars that, when you hear them, make a lot of intuitive sense.
The first is a panel of roughly 150,000 real consumers, predominantly in the US, recruited with detailed demographic data and, critically, purchase receipt history. Extuitive turned these real people into digital personas that can be queried, tested against ads, and sliced any number of ways.
Want to know how moms in Minnesota with a specific vehicle type will respond to your product? You can ask. The key differentiator is that these personas are grounded in what people actually bought, not just what they said they would buy.
The second pillar is brand-specific platform data. When a brand connects their Meta or TikTok account to Extuitive’s platform, the system ingests their full ad history (every metric from click-through rate all the way down to CPA and ROAS) and builds a bespoke model that reflects the specific audience dynamics of that brand. Because even two brands selling similar products can have meaningfully different customer profiles, and a generic model won’t capture that.
The fusion of these two inputs produces what Armen calls a custom fusion model. One that’s rebuilt from scratch for every brand they work with.
The Practical Payoff: Pre-Validation
If you’re a brand that typically launches 200 ads per week, Extuitive’s platform can tell you, before you spend a dollar, that, for example, only 65 of those are worth launching. Of those 65, roughly 50 are likely to perform very well. The other 135?
Don’t bother.
You just saved the time, the creative resources, and the testing budget you would have burned figuring that out the hard way.
Armen also mentioned a third use case that I hadn’t initially anticipated: audience discovery. The platform can identify new consumer segments that the brand has never targeted before but that the data suggests should be responding to their products. Which moves the whole concept beyond efficiency into genuine growth opportunity.
The Results Speak for Themselves
Armen walked through a case study involving a well-known supplements brand. Before Extuitive, they were spending roughly $50,000 per week on ads across a major platform, launching 50 to 70 ads weekly, and finding that only seven or eight performed well, with roughly 20% of their budget going to ads that produced essentially zero conversions.
After Extuitive identified two new audience personas and created targeted creative for them, the results were significant. Approximately 2x improvement in average order value, 2.7x increase in ROAS, 70% improvement in click-through rate, and 60% improvement in conversion rate.
What It Takes to Get Started
For brands considering this kind of approach, whether with Extuitive or anyone pursuing a similar methodology, Armen was clear about what the prerequisites look like. You need at least six months of ad history on Meta or TikTok, and a minimum of roughly 250 ads in your account history to give the model enough signal to work with. The calibration period is 30 days, there’s no charge during that window, and Armen said the bespoke model can be built and quality-checked within about 48 hours of connecting your account.
In terms of company size, Extuitive works with brands ranging from $5 million in topline revenue up through the hundreds of millions, and is in conversations with companies in the tens of billions. The size isn’t the gating factor; it’s the depth of your existing ad data.
The Bottom Line
If I step back from this conversation and think about what Armen really gave us, it’s a framework. The problem of not knowing which ads work isn’t going away. In fact, as AI-generated content floods every platform and ad costs keep climbing, solving it is becoming a genuine competitive differentiator. The brands that figure out how to be more precise, more predictive, and more efficient with their ad spend are going to have a meaningful structural advantage over those still relying on gut feel and platform optimization.
John Wanamaker didn’t have the tools to solve his problem. In 2026, the tools exist. The question is whether you’re using them.
To hear the full conversation with Armen Mkrtchyan, listen to or watch this episode of the Omni Talk Retail Technology Spotlight wherever you get your podcasts:
Apple Podcasts | Spotify | SoundCloud | Amazon Music | YouTube
Be careful out there,
– Chris and the Omni Talk team
P.S. See our past 8 years of wonderful Spotlight Series podcast guests, featuring roughly 200 movers and shakers in retail, by clicking here
Music by hooksounds.com
*Sponsored Content*
Omni Talk® is the retail blog for retailers, written by retailers. Chris Walton founded Omni Talk® in 2017 and have quickly turned it into one of the fastest growing blogs in retail.