Mutinex and Trade Desk Pilot Incrementality-Driven Programmatic Bidding with Hershey

Mutinex (a marketing mix modeling company), The Trade Desk (the largest independent programmatic ad-buying platform), and Hershey ran a pilot that wired incrementality data directly into programmatic bidding. Instead of chasing cheap impressions, the system bids toward ads that demonstrably cause a purchase. Mutinex CEO Henry Innis says brands that don't move this direction within six to twelve months face a structural competitive disadvantage.

The idea is real. The timeline is marketing.

Why The Trade Desk wants this story

AppLovin's performance pitch has been eating the open-internet buy-side narrative, and The Trade Desk's answer has to be that the open web actually converts. Incrementality-gated bidding is that answer. For measurement vendors like VideoAmp, iSpot, and Nielsen, there's a problem embedded in this: if brands want incrementality piped directly into the DSP, MMM specialists like Mutinex slide in between the measurement layer and the bidder. Whoever owns that wire owns a chokepoint everyone else has to route through.

Mutinex is small and already has the integration built. The Trade Desk has a clear motive to make this proprietary rather than an open plug-in any DSP can adopt. Leaving it open hands the same capability to rivals, Google, and Meta, which is the commoditization The Trade Desk is trying to escape.

Where this actually breaks

Hershey got model refresh cycles down to a few weeks. That's fast for an MMM. It's still too slow to steer bids in flight. For the loop to work, you need clean incrementality signals refreshing fast enough to matter, a data-science team that owns the pipeline, and a brand culture willing to let the model cut spend mid-campaign when the trade calendar says push. That last condition is the wall. Almost no CPG organization will let a model turn off spend while the CMO is watching a campaign run.

The Hershey case worked partly because $2 billion in annual spend justifies a dedicated team. Note also that the $2 billion includes trade spend, the money paid to retailers for shelf placement and promotion, which is where the real prize is: closing the loop on offline conversions, not shaving open-web CPMs. A brand running $20 million cannot staff this. The cost of owning the loop swamps the incremental gain from smarter bids. This is a capability for the largest fifty or so advertisers. Most of the market won't see the payback.

Winners and losers

Large advertisers with in-house marketing science already staffed gain a real performance edge if they commit the integration work. For everyone else, the economics don't close.

Traditional measurement vendors face genuine displacement pressure. If incrementality flows directly into the DSP via an MMM vendor, the incumbent measurement layer loses its position in the buying loop.

Our call: By the June 2027 Cannes Lions, The Trade Desk will have locked up an incrementality and MMM vendor through a formal exclusive partnership or an acquisition, with Mutinex the most likely target, making outcomes-based bidding a proprietary Trade Desk feature rather than an open plug-in. The pull toward owning the wedge is stronger than any pull toward openness. Confidence is medium: the motive is clear and the target is small, but timing depends on pilot results and Trade Desk's roadmap.

Before any brand commits to this direction, run the ugly case first: put the loop live and watch what happens the first time the model says stop during a campaign the CMO cares about. That moment tells you whether you have infrastructure or a demo.

Your draft

Mutinex, The Trade Desk, and Hershey wired incrementality data (a measure of whether an ad actually caused a sale, not just showed up next to one) straight into programmatic bidding, so bids chase real business value instead of the cheapest thousand impressions. Mutinex CEO Henry Innis says brands that don't do this within six to 12 months face a structural competitive disadvantage. It's a real idea with a real problem underneath.

What's being decided: whether "outcomes-based bidding" is a product operators should build toward now, or a Cannes demo that runs into brand-org reality. Easy to undo: this is a pilot, nobody has committed a roadmap to it. The deadline is set by Innis's own six-to-12-month clock, and by whether Google and Meta build the same thing for their own inventory first.

The Market Analyst. The Trade Desk needs this story. AppLovin's performance pitch has been eating the buy-side narrative, and Jeff Green's answer has to be "the open internet actually converts." Incrementality-gated bidding is that answer. For measurement vendors it's a threat: if brands want incrementality piped into bidding, VideoAmp, iSpot, and Nielsen either expose the same API plumbing or watch MMM specialists like Mutinex slide in between them and the DSP. Mutinex is small and acquirable, and The Trade Desk has a clear motive to make this proprietary rather than something any vendor can plug into. In plain terms: whoever owns the wire between measurement and bidding owns a chokepoint everyone else has to route through.

The Skeptic. Cannes Lions is where pilots go to become press releases. Hershey got cycle times down to "a few weeks." That's fast for a marketing mix model, still too slow to steer bids in flight. Three things have to be true for this to scale, and none are yet: clean incrementality signals refreshing fast enough to matter, brand data-science teams able to own the loop, and a brand culture that accepts the model saying "stop spending" mid-campaign. That last one is the wall. Almost no CPG organization will let a model turn off spend when the trade calendar says push. Innis's structural disadvantage is real in theory and a five-year problem in practice.

The Operator. The thing that breaks first isn't strategy, it's the pipeline. Someone has to own the schema, the refresh cadence, and the fight that happens when the model says bid down and the campaign manager says we're behind on delivery. That person doesn't exist yet at most brands. At 60 days, media planners and programmatic traders start a turf war over who controls the incrementality threshold. At 90, it's data governance and model versioning fire drills. The Hershey case worked partly because $2 billion in spend justifies a dedicated team. A brand running $20 million cannot staff this, and that's most brands.

The CFO. The $2 billion scope is doing a lot of the selling here, and it includes trade spend, which is the money Hershey pays retailers for shelf and promotion. That's the actual prize: closing the loop on offline conversion and trade dollars, not shaving open-web CPMs. But the cost to stand this up is a standing marketing-science function plus an MMM vendor plus DSP integration work. At Hershey's scale the payback math is plausible. Below the top tier of advertisers, the cost of owning the loop swamps the incremental gain from smarter bids. This is a capability for the largest 50 advertisers. Most of the market won't see the payback.

Where they disagree. The Analyst sees a chokepoint worth owning and a likely acquisition. The Skeptic sees a five-year cultural problem that Innis is calling a six-month deadline. They're both right about different layers: the plumbing is buildable fast, the org change is not. The second split is scope. The Strategist read in the briefing wants this to be horizontal infrastructure across every vertical. The CFO says the economics only work for advertisers big enough to fund a dedicated team, which keeps it narrow for years.

What it hinges on. One belief: can incrementality signals refresh fast enough, and can brand orgs tolerate a model that cuts spend, so that bidding actually changes behavior rather than producing a nicer slide? The council leans skeptical on the timeline and bullish on the direction. The idea is right. Innis's clock is marketing. Before committing, a brand should test the ugly case: run the loop live and watch what happens the first time the model says stop during a campaign the CMO cares about. That single moment tells you whether you have infrastructure or a demo.

Prediction: By the June 2027 Cannes Lions, The Trade Desk will have deepened its tie to an incrementality/MMM vendor (a formal exclusive partnership or an acquisition, Mutinex the most likely target) to make outcomes-based bidding a proprietary Trade Desk feature rather than an open plug-in available to every DSP.

Confidence: Medium — clear motive and a small target, but timing depends on The Trade Desk's roadmap and pilot results.

Why: The Trade Desk is losing the buy-side performance narrative to AppLovin, and "the open internet actually converts" is the counter it needs, which means incrementality-gated bidding has strategic value only if it's differentiated rather than something any competitor can wire in. Mutinex is small enough to buy and has already built the integration, so the cheapest way to own the wedge is to lock it up. The opposite outcome, The Trade Desk leaving this as an open standard any DSP can adopt, hands the advantage to rivals and Google and Meta, which is exactly the commoditization Green is trying to escape. The pull toward capturing the loop is stronger than the pull toward openness.

Revisit by 2027-06-30: We're right if The Trade Desk announces an acquisition or an exclusive/proprietary incrementality-bidding partnership (Mutinex or a comparable MMM vendor) by the June 2027 Cannes Lions. We're wrong if outcomes-based bidding ships as an open integration any DSP can use, or if The Trade Desk makes no such move by then.