The AI Advertising Podcast: S2
Episode 9
AI in Retail: Relevance at Scale and Incremental Growth
About This Episode
As AI reshapes retail advertising with real-time optimization, faster creative testing, and new ways to reach shoppers across CTV, the open web, and commerce environments—one big question remains:
Can AI actually deliver relevance at scale and prove what’s truly incremental?
Dana Karlin | Head of Commerce, StackAdapt
Transcript
Diego Pineda (00:00:00)
Retail marketers are feeling the squeeze. Budgets are tighter. Reporting doesn’t always match across platforms. And leadership keeps coming back to one blunt question: did this drive new sales, or just take credit for sales that were already going to happen?
The industry is spending fast. WARC expects global retail media spend to hit $196 billion dollars in 2026—about 16% of all ad spend.
Meanwhile, product discovery is changing. Adobe reported generative AI traffic to U.S. retail sites was up 4,700% YoY in July 2025.
So how do retail and ecommerce marketers win when shopping journeys happen everywhere, not in a straight line? And how do you prove what’s truly working?
Today I’m joined by Dana Karlin, Head of Commerce at StackAdapt—previously part of the team that helped build TikTok’s retail vertical.
We’re going to unpack what’s stressing retail marketers right now, how to make relevance actionable across channels, and the real impact AI is having on commerce marketing.
Podcast Intro (00:01:25)
Welcome to the AI Advertising Podcast, brought to you by StackAdapt. I’m your host, Diego Pineda. Get ready to dive into AI, Ads, and Aha moments.
Diego Pineda (00:01:40)
Retail marketers are under pressure to grow sales and show clear ROI.
Dana Karlin (00:01:45)
Q1 is supposed to be our break, but it feels like lately within the last few years, we haven’t had much of a break. So I would say the biggest thing that are keeping people up at night is the pressure to prove incrementality. We have shrinking budgets and really fragmented consumer journeys. So retailers are being asked to do more with less, while shoppers are moving more fluidly than ever across channels. They’re discovering on social, they’re researching on CTV, they’re converting on retail media networks. So the funnel hasn’t necessarily disappeared, but it’s collapsed. And shoppers are not necessarily moving in stages anymore. They’re really moving in moments.
Diego Pineda (00:02:30)
When shoppers move in moments, it changes the boardroom question from “what performed?” to “what actually drove new demand?”
Dana Karlin (00:02:40)
The biggest question that we get is how we’re generating demand and not just capturing what’s already there. CFOs and financial folks want to see more testing than ever. They want to see lift studies. They want to see incrementality testing. And they want proof that media spend is actually generating demand and not just intercepting existing intent. The challenge is especially acute for e-commerce brands specifically that are over-reliant on search and bottom funnel tactics. Search is capturing demand. It’s not necessarily creating it.
Diego Pineda (00:03:20)
So, if you need to generate demand, then you can’t just pour everything into the bottom of the funnel, and hope for the best.
Dana Karlin (00:03:33)
Three key places are where we see spend wasted. Overinvestment in bottle bottom funnel tactics without upper funnel support. That’s something unfortunately we see a lot. um This leads brands to harvest demand but not create it as I’ve said, which is super important. um And then that’s immediately followed by a media spend reaching diminishing return. So I would also say number two is channel silos that create redundant targeting and frequency waste across retail media networks, programmatic and social. And then finally, the third thing I would say is treating all first party data equally. so recency and context matter enormously. Automating noise is easy. But automating relevance is the real challenge.
Diego Pineda (00:04:23)
That last thought—automating relevance vs automating noise—is the entire game. But to do it, we have to understand where decisions are starting to form. This is a topic we discussed in a previous episode with Becky Tasker, where we talked about filling the upper funnel so the bottom funnel doesn’t dry up.
Now, it is said that retail is everywhere now. But “everywhere” doesn’t mean “spray and pray.” I asked Dana what she thinks this means.
Dana Karlin (00:04:54)
I would say research and validation moments that happen away from point of sale. So think product review, comparison shopping on publisher sites, social proof from creators, contextual moments like recipe content for CPG or home improvement articles for furniture. for furniture if you don’t own the checkout, you have to own the moment is something that I always like to say.
At StackAdapt, we see shoppers engaging with contextual content showing a much stronger purchase intent than those that are just browsing retail purchase categories.
Diego Pineda (00:05:29)
Those moments are spread across retail sites, the open web, social, and streaming. Each platform measures success differently. Privacy rules and signal loss make it harder to connect the dots. So marketers end up with a bunch of separate reports—and no single story.
That’s why measurement standards have become such a big deal. Groups like the Interactive Advertising Bureau (IAB) and the Media Rating Council (MRC) are pushing for clearer, more consistent ways to define and measure retail media across onsite, offsite, and even in-store.
So if the shopper journey is scattered—and the measurement is scattered—what’s the system that can connect signals to activation, and activation to outcomes?
Diego Pineda (00:06:23)
This is where the idea of relevance at scale really matters. By that, we mean showing the right message to the right person at the right time—across all the places people discover and shop. And the argument is: AI helps you figure out what “right” looks like in real time, but you still need a system to activate that relevance across the open web, CTV, and retail environments—and then measure what actually drove new sales. Dana breaks this down with a concrete commerce example.
Dana Karlin (00:06:58)
So let’s say a shopper abandons a cart for running shoes. That signal gets ingested into our platform within minutes. Our AI identifies intent earlier by evaluating that commerce signal alongside contextual predictive signals. ah That serves personalized creative while bidding higher when that user appears on premium fitness content. So AI thinks in moments, programmatic moves in real time. So within 48 to 72 hours, we can see conversion lift and feed that signal directly into optimized bidding and creative in real time.
Diego Pineda (00:07:36)
That’s the loop: signals, context, creative, bidding, outcomes—then learning. So where is the real impact right now?
Dana Karlin (00:07:46)
Bidding for sure. That’s something that really matters and AI is having a big impact because commerce moments are multiplying and machine learning models can process thousands of signals per impression. So adjusting bidding base on time of day, device, contextual targeting, proximity to purchase signals and inventory. At StackAdapt, our bidding algorithms consistently learn, which combines predictive conversion and adjust in real time. So we’re actually seeing 20 to 30% efficiency gains.
Diego Pineda (00:08:22)
And it’s not just about optimizing within one “screen.” It’s about orchestrating across the whole journey.
Dana Karlin (00:08:29)
Programmatic provides scale and AI provides the ability to have intelligence. So human planning can’t keep up with the number of moments that are happening across the fragmented consumer journey. um And really what programmatic does is it gives the ability for AI to decide when relevance matters and programmatic comes in and decides where your brand should be showing up. um So we can make bid decisions in mid milliseconds across display video, native, audio, and CTV simultaneously versus obviously social is just one screen.
Diego Pineda (00:09:08)
This is where AI stops being a headline and starts showing up in innovative retail marketing strategies. People are using new tools to discover products, validate choices, and narrow down what they want. So the number of moments that shape a purchase keeps growing. And marketers have to stay relevant in all of them. Which is why creative—and how fast you can adapt it—matters more than ever.
Dana Karlin (00:09:35)
Dynamic creative at true scale is something that’s been unlocked recently and is changing the industry. So we can now produce hundreds of variations and test them all simultaneously. AI enables creative to adapt content, intent, and environmental issues in real time. A product ad next to a recipe article highlights cooking features while the same product on a fitness site emphasizes health benefits. So AI turns creative from a static asset into a dynamic experience and programmatic delivers the right creative at the right moment.
Diego Pineda (00:10:14)
But with generative creative comes risk—brand safety, weird outputs, pushing brands beyond comfort zones.
Dana Karlin (00:10:25)
When we first launched AI creative at my previous employer, we had a lot of issues with the generative AI creative. And there were some things that were launched that brands were not necessarily comfortable with. It’s very much pushing brands outside of their comfort zone and what has worked in the past. So I think it’s something that we’re all figuring out together and it’s something that we all really need to like to lean into testing together. It’s new to consumers as well. So I think it’s something that… keeping brand authenticity at the center of everything that we’re doing is absolutely crucial so that brands can remain authentic, but also bring users along. And we all understand that AI is the center of conversation right now. Like if you watch the Super Bowl, what percentage of the ads were AI related. So humans understand that this is coming, that it’s going to be a big part of their life. So I don’t think that we’re going to see as high of aversion to it as I think we originally guessed. But that testing framework is just so crucial. And AI truly gives us the ability to optimize in ways that we never have before. So it’s definitely a positive for me.
Diego Pineda (00:11:47)
When commerce is everywhere, relevance isn’t optional. AI makes relevance possible at scale. Programmatic makes it actionable—across channels and in real time—a proof layer. Dana defines “relevance at scale” in a way that connects directly to incrementality.
Dana Karlin (00:12:05)
So talking about relevance at scale, to me, that means that we’re delivering the right message to the right shopper at the right moment across every channel that they use with measurement that proves it drove incremental purchases, not just capturing existing demand.
Diego Pineda (00:12:21)
Now, two practical questions retail marketers should always ask are: How do we prove incrementality quickly? And, what’s the one move we should make this quarter?
Dana Karlin (00:12:34)
I think brands being able to prove incrementality is becoming BAU, right? And it’s something that we must do consistently. So I would say geo holdout testing or audience-based holdout testing is where we can truly prove incrementality. So take 15%, 10% maybe of your audience, completely suppress them from your campaign, and then compare conversion rates and purchase frequency between the exposed and holdout groups over, say, 30 to 45 days. Pair that with multi-touch attribution that goes beyond last click and credits upper funnel and mid funnel touch points. For example, at StackAdapt, we help clients set these tests up so that they can measure lift across both direct conversion and assisted conversion. The goal in 60 days isn’t perfect attribution, but it’s credible evidence of incremental impact.
If I were to give a retail leader one piece of advice this quarter, it would be to start layering contextual targeting on top of your audience targeting and then measure the lift.
Take your high performing audience segments and test them in contextually relevant environments. home improvement content for home goods, lifestyle and wellness for beauty, recipe content for CPGs, you’ll likely see higher engagement, better conversion rates and stronger brand lift measurable within six to eight weeks. When commerce is everywhere, relevance isn’t optional. AI and programmatic make it actionable.
Diego Pineda (00:14:10)
Here are the practical takeaways for your next planning cycle:
First, anchor your strategy in demand creation—not just bottom-funnel capture.
Prove what’s real: run a simple, defensible incrementality test (geo or audience holdout) over 30–45 days.
Treat signals like signals, not noise: prioritize recency and context, not just “more data.”
Let AI do what it’s best at today: real-time bidding and optimization across thousands of signals per impression.
Build creative for automated relevance: keep it modular, test lots of variations, and use clear brand guardrails so personalization stays on-brand.
And, finally, win outside the checkout: show up in the research and validation moments, especially if you don’t own the point of sale.
In a world where retail is everywhere, the brands that win are the most relevant, in the moments that matter, with proof to back it up.
Podcast Outro (00:15:23)
Thanks for listening to this episode of The AI Advertising Podcast. This podcast is produced by StackAdapt. Visit us at stackadpat.com for more information about using AI in your advertising campaigns. If you liked what you heard, remember to subscribe, and we’ll see you next time.
You May Also Like
Report
GLOBAL
Feb 5, 2026
Personalization in Digital Marketing: Trends, Statistics, and What’s Next
