Your Marketing Reports Are Already Wrong. Hershey Just Fixed That With AI.
May 25, 2026 | 4 min read
Here is a sentence that should make every CMO uncomfortable.
“We were getting the full read of 2024 midway through 2025, while we were planning for 2026.”
That is Vinny Rinaldi, VP of Media and Marketing Technology at Hershey, describing what marketing measurement looked like before they changed everything. A $2 billion annual marketing operation, optimizing on data that was 12 to 18 months old by the time anyone could act on it.
If you are still running quarterly marketing mix modeling, your situation is not that different.
The Problem Nobody Talks About
Marketing measurement has had a dirty secret for decades. By the time your MMM results come back, you have already committed next year’s budget, briefed your agency, and locked your creative calendar. The analysis tells you what worked last year. It has almost nothing to say about what you should do next month.
Hershey’s Rinaldi put it more precisely than most: they were receiving the full picture of 2024 while actively planning for 2026. Two full planning cycles behind the data. Every optimization they made was based on incomplete or irrelevant information.
This is not a small inefficiency. On a $2 billion spend, even a 2 percent misallocation is $40 million. The waste is structural, and it has been hiding in plain sight because everyone assumed it was simply the cost of running a large marketing operation.
It is not. It is the cost of slow data.
What Hershey Actually Built
The company deployed two platforms in combination: Mutinex for AI-driven marketing mix modeling, and Tracer for cleaning and standardizing the underlying data infrastructure.
Mutinex is built as a multi-agent system, where each agent functions as a domain specialist. One agent handles marketing econometrics. Another applies competitive pricing theory. A third monitors for model failure and flags when the analysis is drifting from reality. They run in parallel, not in sequence. That is why what used to take months now completes in three weeks.
Mutinex runs on Claude and Gemini. The models are not running the analysis end to end. They are coordinating specialists that each hold a distinct piece of the measurement problem. The architecture matters because it means Hershey is not getting a single black-box output. It is getting a system of checks that can explain its own conclusions.
Tracer handles what Mutinex cannot: the messy, inconsistent, siloed marketing data that most large brands have accumulated over decades. Naming conventions that do not match. Attribution windows that conflict. Spend data that lives in six different systems. Tracer cleans all of that before Mutinex ever sees it. That is the unglamorous part. It is also the part that makes everything else possible.
> The result: Hershey’s entire brand portfolio, analyzed every month. Twelve reads per year instead of two or three.
What $80 to $100 Million Looks Like
Hershey expects a 4 to 5 percent lift in media-attributable revenue from this system.
On a $2 billion spend, that is $80 to $100 million in recovered value. Not from a new channel. Not from a better creative. From knowing which channels are working while there is still time to reallocate budget to them.
That number reframes what marketing technology is actually worth. The conversation inside most organizations treats measurement as a cost center: a reporting function that explains outcomes after the fact. What Hershey has built treats measurement as a revenue driver: a system that generates value by accelerating the speed at which decisions can be made.
The return on investment case for real-time MMM is not complicated. The gap between when you spend and when you find out if it worked is costing you money every month. The faster you close that gap, the more of your budget goes toward things that actually work.
What to Do This Week
First, audit your current measurement cadence honestly. How often are you running marketing mix modeling? If the answer is quarterly or less, calculate the lag between campaign spend and when your team receives actionable results. That number is your exposure. Most marketing leaders do not know what it is.
Second, map your data infrastructure before you evaluate any AI measurement tool. Mutinex works because Tracer cleaned the data first. If your spend data lives in multiple disconnected systems with inconsistent naming conventions, no measurement AI will save you. Fix the foundation first.
Third, reframe the MMM conversation with your CFO. The argument that wins budget for real-time measurement is not “we need better reporting.” It is “our current reporting is 12 to 18 months behind the decisions we are making today, and that lag has a dollar cost we can calculate.” Bring that number into the room.
The brands that move to near-real-time measurement in 2026 will build a compounding advantage. Every month of faster data is another month of better allocation decisions. Every better allocation decision improves the baseline for the next model run.
Everyone else will be optimizing 2026 with 2024 data. And they will not realize the cost until it shows up in 2027’s results.
What is your current measurement lag? How many months behind are you when you receive your MMM results? Drop it in the comments.
—
#AI #Marketing #DigitalMarketing #AIMarketing #MarketingMixModeling #MMM #MediaMeasurement #Hershey #Mutinex #AIAgents #MarTech #PerformanceMarketing #MediaStrategy #DataDrivenMarketing #MarketingAnalytics #CMO #MarketingLeadership #BrandMarketing #LinkedInMarketing #FutureOfMarketing
