Liquid Death’s Invisible Sales Break AI Models & Ad-Tech

The Invisible Checkout: Why Liquid Death’s Real Shelf Life Breaks Every AI Model

If you crack open a can of Liquid Death in the office, someone will ask where you bought it. The answer is almost always “the bodega on the corner,” not “I clicked a shoppable ad.” That’s the joke the industry keeps missing: the fastest-growing non-alcoholic beverage in America is functionally invisible to the dashboards we now trust to steer billions in media. Roughly 80 % of Liquid Death’s revenue walks out of physical retail locations with no digital receipt, no loyalty swipe, and no persistent ID. Your algorithm can’t optimize what it can’t see, and right now it can’t see the register tape.

The AI Buying Bench-Press: Whoever Trains the Model First, Writes the Rules

While marketers lament the blind spot, sell-side players are quietly building new walled gardens inside black-box AI. Last week SSP Kargo pulled back the curtain on Project Kera, a closed-beta, AI-driven unified buying platform that ingests first-party supply, crunches engagement signals, and spits out auto-optimized bids. Early tester Wpromote is already feeding campaign data into Kera’s neural net—effectively tutoring the algorithm on what a “quality” impression is worth.

Here’s the punch-line: in machine-learning auctions, the first trainer becomes the de-facto pricing ruler. Every subsequent buyer inherits the bid-floor taxonomy baked in during beta. If Wpromote’s KPIs favor viewability over click-through, the model encodes that bias and charges the next buyer accordingly. Early testers don’t just get cheaper CPMs—they get to write the math everyone else pays rent on. Expect rival agencies to bang on Kargo’s door, but the rent’s already going up.

Linkby’s $15 Million Bet: Turning Publisher Content Into a First-Party Flywheel

The same day Kera slipped into beta, Linkby announced a $15 million Series B to scale its “reader-engagement” native marketplace. Instead of impressions, brands pay when a reader scrolls past a certain depth or clicks a contextually placed product card. Every interaction is captured against the publisher’s first-party ID, then piped back to the advertiser as a deterministic seed for look-alike modeling.

Translation: even mid-tier publishers can now monetize their own audiences without cookies, while brands collect consented data that actually survives Safari and Chrome deprecation. It’s not a replacement for the missing Liquid Death POS signal, but it’s another proof point that capital is sprinting toward anything labeled AI-driven first-party and leaving cookie-alternative startups gasping for air.

Holding-Company Chess: Why Publicis Dumped The Trade Desk—and Why Viant Is Sending Flowers

The other seismic shift happened inside the conference rooms of Publicis Media. Two weeks ago the holding company quietly stopped recommending The Trade Desk to clients, a move that shifted hundreds of millions in programmatic spend. Rival DSPs immediately flooded the inbox, with Viant described by one buyer as “the most aggressive,” flying C-suite execs to client rooftops for sunset cocktails.

But this isn’t a simple pricing tantrum. Publicis has spent the last three years stitching Epsilon’s identity graph to its own Kinetic DSP and now to supply-side tools like Kera. The goal: swap low-margin media arbitrage for high-margin data-as-a-service. By black-listing TTD, Publicis nudges clients into environments where it controls the fee stack—from clean-room onboarding to bid-log analytics. Every dollar that exits TTD potentially lands in a Publicis-owned SaaS line item with 80 % gross margins. The holding company isn’t picking a fight; it’s re-balancing the balance sheet.

Benchmarking the Unmeasurable: What Happens When the Cash Register Doesn’t Phone Home

Put these threads side-by-side and the gag becomes clear: the industry’s newest currency—AI-trained first-party data—still can’t buy a single SKU at the physical checkout. Liquid Death will likely sell another 200 million cans this year, but most will vanish into the same measurement void that plagued CPG brands in 1998.

Some startups are trying to bridge the gap. Bridg, Fetch, and Upside all dangle receipt-scanning or card-linked offers to re-attach SKU data to a device graph. Yet adoption is patchy, and no retailer has agreed to open its UPC fire-hose at scale. Meanwhile, Amazon’s Just Walk Out tech remains locked inside Whole Foods and ballparks—hardly the corner bodega.

So marketers do what they always do: benchmark against the metric they can measure (digital ROAS) and pray it correlates to the metric they can’t (incremental retail sales). The comic tragedy? The faster Liquid Death grows, the worse that proxy performs.

Fade-Out: Audit the Cash Register Before You Audit the Algorithm

Venture dollars don’t lie. In Q1, 62 % of ad-tech VC funding carried the tag “AI-first-party,” up from 11 % tagged “cookie-alternative” a year ago, according to CB Insights. Every check writer has internalized the punch-line: the cookie is dead, long live the algorithm.

But until someone solves the offline attribution equation, every AI model is optimizing toward a proxy that may have zero correlation to real-world sales. If you can’t benchmark the cash register, you can’t benchmark the campaign. And that, dear reader, is the only metric that matters—whether you’re selling water, widgets, or whatever DSP is courting your budget this week.