The gap between where commerce is headed and how most brands are operating has never been wider. Our team attended this year's Commerce Media Brand Summit, and that tension ran through every session. Here's what you need to know.

The Measurement Problem Is Still the Problem

Let's just get this out of the way, because it dominated nearly every session.

A moderator on one panel asked the room to raise their hands if they were at least 50% satisfied with their current measurement capabilities.

Almost no one raised their hand.

This isn't a new observation. We've been writing about measurement fragmentation for years. What stood out this time was the exhaustion in the room. Brands aren't confused about what good measurement looks like; they're stuck. They know ROAS is a flawed metric. They know incrementality testing gives a better picture. And yet they still can't fully act on it, because the data comes in six to twelve weeks late, platforms don't share it consistently, and organizational incentives reward channel-specific wins over total business performance.

One brand leader put it plainly: when 80% of your sales happen in-store, and you're measuring digital campaigns, everything is modeled. Her analogy — that it's like an AI generating an image that slowly resolves into an animal after 100 iterations — landed because it's accurate. And her workaround wasn't glamorous: POS data, share position, and incremental distribution. Not AI-powered. But real and fast-moving.

What this means for brands: Stop waiting for the perfect measurement solution. It's not arriving in the next 18 months. The brands winning right now are making directional decisions with imperfect data rather than waiting for clean attribution that doesn't exist. The goal is to stay within guardrails, not hit a perfect number.

Retail Media Is Maturing, and the Easy Gains Are Gone

Retail media is working, but the growth is getting harder not because budgets are shrinking, but because the low-hanging fruit is gone.

That's the shift from a growing channel to a mature one. Spending more doesn't automatically produce the same returns anymore. Operating differently is now the requirement.

Survey data from the summit showed retail media budget allocation trending upward across almost every category, averaging around 15% of total media spend, with personal care leading at nearly 29%. Food and beverage jumped 5 points year-over-year, signaling that even traditionally cautious CPG companies are committing serious dollars. The one outlier: alcohol, which is declining due to a combination of industry headwinds and a three-tier distribution system that makes digital commerce genuinely complicated.

By 2027, retail media is projected to become the third-largest digital media channel behind search and social, a prediction that's tracking on schedule.

The catch: brands are still underinvesting relative to the opportunity, and the reason is always measurement. Retailers need to share more data, faster. The brands already investing are doing so on directional signals, which means the brands waiting for perfect data are simply ceding ground.

What this means for brands: The question is no longer whether to invest in retail media. It's how to operate it differently as it matures. That means moving beyond on-site search toward full-funnel approaches, off-site, display, connected TV, digital out-of-home, and building the internal capability to measure across those touchpoints, even imperfectly.

The Funnel Isn't Broken. Your Org Chart Is.

This came up in at least three sessions, and it's probably the most actionable takeaway from the whole summit.

One speaker shared a case study: a health and beauty brand running Amazon campaigns alongside TikTok campaigns, except they actually planned them together shared creative, shared targeting signals, shared measurement. When TikTok ran in coordination with Amazon, Amazon's ROAS improved 10%. TikTok didn't just drive awareness. It made Amazon convert better.

Most brands would never see that lift, because their TikTok team and their Amazon team don't have shared planning meetings. Data stays siloed. TikTok looks inefficient. Amazon looks like the hero. Budgets get allocated accordingly, which is exactly backwards.

Another executive reorganized her entire team from channel ownership to customer ownership. Instead of a social team, a search team, and an Amazon team fighting over attribution, she built around shared brand contribution margin. Channel P&Ls still exist, but at the CMO level, they're directional. The rollup is what matters.

Jack Kniper from Purdue Farms put it well: "Everyone is goaled by their individual metrics, but there's no shared goal." He described a retail search campaign that mysteriously spiked and when he dug in, an organic video was going viral and driving the demand. Two separate teams, no shared planning, pure luck that he even connected the dots.

What this means for brands: Start with the meeting that doesn't exist. Get your Amazon team and your social team in the same room, with the same goals, at least monthly. You don't need a full reorg to start. You need a shared north star metric — total incrementality, overall brand contribution — and the willingness to treat channel-level numbers as directional rather than gospel.

AI Shopping Is Real, But Agentic Checkout Is Further Away Than Expected

This was the most nuanced conversation of the summit, and the data from Criteo was genuinely useful.

In a survey of roughly 6,400 consumers globally, 61% said they're somewhat or very likely to use AI to help make a purchase decision in the next year. That's a big number. But when it gets to the actual checkout moment, having AI complete a purchase on their behalf, that comfort drops to about 30% of Americans. Globally, it's lower.

The timing tells: agentic shopping, where you instruct a system to go buy something for you, has existed for a decade. The summit room couldn't name more than one or two people who'd actually used it. The gap between capability and consumer trust is measured in years, not quarters. eMarketer data suggests only about 2% of e-commerce checkouts will be automated by agents by 2029 — and that estimate came before OpenAI quietly pivoted away from in-environment checkout.

What AI is actually doing well right now: discovery and comparison. One attendee described using ChatGPT to research a hair tool before buying on Amazon. It's a research layer, not a transaction layer. The shopper searched in ChatGPT, validated in customer reviews, and purchased on Amazon — three platforms, one purchase, no clear attribution trail anywhere.

Criteo is addressing this through an MCP-accessible data layer that gives LLMs access to real-time retailer data. They reported a 60% improvement in product recommendation accuracy when that layer is in place — and they're now OpenAI's first commerce data partner. The infrastructure of AI shopping is being built right now, and the brands that understand what's powering those recommendations will have a real advantage.

What this means for brands: Your content and product data strategy is now your AI discoverability strategy. Optimized PDPs, complete product attributes, accurate pricing, and clear use-case descriptions aren't just SEO hygiene anymore. They're the inputs that determine whether an AI system recommends you or your competitor.

The Emotional Relevance Thread Nobody Is Talking About Enough

The opening keynote was about something that sounds soft until you see how strategically rigorous it actually is: emotional relevance in commerce.

The argument: the brands winning in commerce media aren't winning because their targeting is better. They're winning because their connection is stronger. Stand out, create emotional impact, then use purchasing data to personalize,  in that order.

The Nestlé example was memorable. Twenty years ago, they failed to sell coffee in Japan. So they researched why Americans love coffee — and found that the scent triggers the same brain region as memories of childhood and home. They responded by flooding Japan with coffee-flavored candy and ice cream. Two decades later, Nestlé is selling coffee to adults who grew up craving it emotionally.

Most brands won't play a twenty-year game. But the principle translates: connection precedes conversion, and emotional relevance makes the conversion feel inevitable rather than persuaded.

The counterintuitive point: with all the focus on data and attribution in every other session, the opening talk was a reminder that Costco customers don't walk out with $187 worth of unplanned items because of targeting. It's because of the relationship. Brands investing only in performance marketing and ignoring brand equity are renting demand, not building it.

The Through Line

Every major theme at the summit circles back to the same tension: the tools to run full-funnel, cross-channel, AI-informed marketing exist. The organizational structures, measurement systems, and budget models to support it mostly don't yet.

The brands making progress stopped waiting for perfect conditions. They're reorganizing around the customer journey instead of channels, using imperfect incrementality data instead of ROAS, and investing in the AI discovery layer before they fully understand how it converts.

The brands that are stalled are waiting for the data to get cleaner, the org chart to get sorted, and the AI landscape to settle. That's understandable. It's also how you end up two years behind.