Murder Your Thirst And Measure Everything

Full analysis

Liquid Death's Benoit Vautere, the brand's chief media and digital commerce officer, spent an hour with AdExchanger's Allison Schiff explaining how a fast-growing CPG brand builds real incrementality measurement when most of its sales happen on a physical shelf. The takeaway for operators: a sophisticated buyer is treating a high ROAS as a warning light, paying out of pocket for zip-code point-of-sale data, and building his own attribution instead of trusting the platforms and retail media networks to grade their own homework.

Reversibility: Type 2 for any single operator (measurement approaches are cheap to test and easy to reverse). But the underlying shift in what advertisers demand is closer to Type 1 for the vendors on the receiving end.

What's actually being decided: not "should Liquid Death buy more CTV." It's whether the measurement stack sold by walled gardens and retail media networks survives contact with a buyer who insists on independent proof of lift. Timeline: no forcing event, this is a slow grind through 2026 budget planning.

The Market Analyst Follow where the skepticism points. Vautere is moving money toward connected TV (ads on streaming) for awareness and away from paid social, which he calls "crack cocaine." That's one brand, but it rhymes with a broader trend: buyers want channels where creative stands out and where they can prove incremental sales, not just harvest demand that would have converted anyway. Measurement vendors that can independently verify lift, and CTV sellers, are the winners in his framing. The loser is any platform whose pitch is "trust our black-box optimization." For a smart generalist: he's spending where he can prove the ad actually caused the sale.

The Skeptic Steelman the doubt about Vautere himself. He's talking his book. He built V-Tags, an attribution company, before joining Liquid Death, so of course he privileges independent measurement over Meta's numbers. His Meta critique is one-sided, and Meta would say Advantage+ automation lifts performance. And his "50% of revenue from light buyers" stat is Byron Sharp lore, directionally right but not sourced to a number. The useful part survives the discount: treating a suspiciously high ROAS as demand harvesting is correct mechanically, whoever says it.

The Operator Try to actually run this Tuesday morning. Vautere buys point-of-sale data from retailers at zip-code level, then normalizes it across chains that all format it differently. That normalization is the whole job, and it's ugly. Then he runs matched-market tests, comparing a region where ads ran against a similar region where they didn't, because he can't run a proper marketing mix model. A standard MMM wants roughly two years of stable history, and a brand growing this fast doesn't have a stable baseline. So he runs shorter "mini-MMMs" and only trusts effects big enough to hold up even with sloppy data. In plain terms: he's given up on precision and hunts only for signals too large to argue with.

The Customer / End User Here the customer is the retail media network. Walmart Connect, Target Roundel, Kroger, Amazon all want to own the full loop: sell the ads, then grade whether the ads worked. Vautere just told them no. He'd rather pay for raw POS data and build his own incrementality proof than accept the retailer's marked-up scorecard. That's the signal operators should sit with. The most sophisticated CPG buyers don't want the walled garden's self-serving measurement, they want the raw material to check it. In plain terms: the referee also owns one of the teams, and the smart buyers noticed.

The CFO Count the real cost. Buying zip-code POS data across Walmart, Target, Kroger and Amazon, normalizing it, and staffing people to run matched-market tests is expensive and only pays off at Liquid Death's scale. A mid-size brand can't afford this and will keep taking the retailer's numbers on faith. So the independent-measurement path is a big-advertiser luxury today. The opportunity that creates: a vendor who can package this workflow, buy the POS data once, and sell normalized incrementality to the brands too small to build it themselves.

The tensions

Two disagreements matter. First, the Skeptic versus the Market Analyst on how much to trust Vautere. He's an attribution founder telling you attribution is underrated, and he's justifying a move away from the channel he dislikes. The method is sound even if the messenger is motivated. Second, the CFO versus the Customer on how far this spreads. Vautere can afford to reject the retailer's scorecard because Liquid Death is big enough to buy the raw data. The retail media networks are betting most brands can't, and stay dependent. That bet is probably right for the long tail and wrong for the top.

Synthesis

This episode hinges on one belief: that the biggest CPG advertisers will keep refusing to let sellers grade their own work, and will pay for independent proof of incremental sales. The council leans toward that being real and durable at the top of the market, and toward it being unaffordable for everyone else for now. The gap between those two is the business opportunity, an independent incrementality layer for retail media that a mid-size brand can actually buy.

What to verify before acting on it: whether zip-code POS data from the major retailers is licensable at a price that works below Liquid Death's scale, and whether matched-market testing holds up in categories with thinner store coverage than bottled water. If both hold, there's a product. If the data licensing stays gated to whales, the retail media networks keep their measurement monopoly by default.

No high-conviction prediction this week.

This is a practitioner master class, not a market event. The strongest claim available, that independent incrementality measurement pressures retail media networks, is a slow structural drift with no dated forcing function I can honestly attach a Medium-or-better call to. Forcing a prediction here would be a hunch dressed as conviction, and that pollutes the scoreboard.