Daily Pulse · March 18, 2026 · 3 signals
The Data That Actually Converts
Purchase-verified data is winning.
The Trade Desk's Advertisers Are Leaving. Amazon DSP Is Where They're Going.
Change
Named agencies publicly defecting from TTD to Amazon DSP
→ Price problems get solved with pricing – data moat problems don't
Why it mattersData moats beat pricing in the DSP competition
$2.9B
TTD 2025 revenue
47%
TTD profit margins
5
Named agencies defecting
Five named agencies – VCCP Media UK, Markacy, CrowdLouder, Go Fish Digital, and Butler/Till – publicly described redirecting programmatic spend away from The Trade Desk. Specific complaints: three account team transitions in one year, mid-contract rate increase threats, features automatically applied without consent, and loss of manual controls. The Digiday piece ran one day after Ad Age reported Publicis's formal recommendation withdrawal following its own FirmDecisions audit.
The pattern matters more than the individual names. These are agencies making a public statement, not quietly moving budgets. That's a different signal: they're communicating to clients, competitors, and future audit committees that TTD's AI automation layer has compliance implications they weren't willing to manage.
Amazon DSP is the named primary beneficiary in every case. The reason isn't pricing – Amazon DSP is not a discount product. It's that Amazon offers Prime Video inventory, verified retail purchase data, and a platform with no structural conflict between its media buying and data businesses. That combination doesn't exist anywhere else in the market.
▲ Wins
Amazon DSP – gaining named agency relationships explicitly attributed to TTD defection, expanding its Prime Video + retail data moat as programmatic share shifts. Direct publishers who benefit as agencies route around DSP intermediaries.
▼ Loses
The Trade Desk – the combination of Publicis withdrawal, named agency defections, and Kokai compliance concerns creates a compounding narrative. Each new agency that moves spend validates the previous ones. The business is profitable and dominant; the trajectory is not.
◆ Pulse Take
TTD's $2.9B revenue and 47% margins describe a dominant business – but the defection pattern describes one losing to structural data advantages. Price problems get solved with pricing; data moat problems don't.
Costco Entered Retail Media. Its Members' Purchase Data Has No Competitor.
Change
Costco launched ML-native ads trained on member purchase data
→ The member data is the moat – and it compounds with every transaction
Why it mattersPurchase-verified targeting competes with reach-based platforms
Q2
Partner beta launch
ML
Purchase-trained targeting
#1
Highest-AOV retail member base
Costco launched Reserved Display on its Costco Velocity platform in partnership with Moloco – ML-trained display ads appearing on Costco.com's homepage and search results, targeting based on actual member purchase behavior. Q2 2026 partner beta, broader rollout later in the year.
The product matters less than the data behind it. Costco members are a demographically specific, purchase-verified cohort: bulk purchasers, verified high AOV, membership renewal rates that indicate sustained spending commitment. Meta targets similar households with modeled lookalike audiences. Google targets them through keyword and behavioral inference. Costco has their actual purchase history.
For brands selling at Costco, the targeting option is qualitatively different from what reach-based platforms offer – not more demographic signals, but verified category purchase behavior from the exact customer segment they're trying to reach. The CPMs will reflect that. So will the budget allocation conversations happening in Q2.
▲ Wins
Costco (captures upper-funnel budget from reach-based platforms while monetizing member data). Moloco – earns a marquee retail client for its commerce media ML platform. Brands selling at Costco that can prove purchase lift through display.
▼ Loses
Meta and Google for upper-funnel consumer brand advertising targeting affluent households. Costco member purchase data is verified at the checkout level in a way neither platform can match with modeled audiences or behavioral inference.
◆ Pulse Take
Costco building a retail media network is not about ad revenue – it's about what member purchase data is worth as a targeting layer. The member data is the structural moat, and it compounds with every transaction.
Social +29%, Commerce +24%, CTV +20%. The Channels Where AI Runs Are the Channels Growing.
Change
AI-optimized channels grew 4-5x faster than human-optimized
→ Budget is following automation – and that is becoming the same as performance
Why it mattersBudget divide is now human-optimized vs. AI-optimized
+29%
Social ad spend YoY
+24%
Commerce media YoY
+20%
CTV ad spend YoY
Q4 2025 ad spend data from Madison & Wall, published in Digiday's Ad Tech Briefing on March 17, shows the three fastest-growing channels year-over-year: social at +29%, commerce media at +24%, CTV at +20%. Traditional channels are not in that tier.
The structural feature shared by the three leaders: AI-automated targeting, optimization, and bidding are the default operating mode, not an opt-in feature. Meta's Advantage+, Amazon's ML-native commerce targeting, and CTV platforms' programmatic auction systems all make decisions that human buyers would have made manually five years ago. The budget is following the automation.
The implication for human media planning roles is direct. The channels losing share are the ones where human judgment still drives the primary optimization variables. The channels gaining share are the ones where that judgment has been systematically replaced. The Q4 data is the scoreboard.
▲ Wins
Meta and TikTok (social), Amazon and Walmart (commerce media), streaming platforms with AI-automated targeting (CTV). Any platform where AI makes the targeting and bidding decisions by default.
▼ Loses
Linear TV, traditional display, and channels where optimization is primarily human-managed. Media buyers whose value proposition is discretionary judgment in channel selection and pacing – the data says that judgment is systematically producing lower returns.
◆ Pulse Take
Budgets are moving to AI-optimized channels at 4-5x the rate of everything else. This isn't budget following performance – it's budget following automation, and the two are becoming the same thing.
3 signals · March 18, 2026