> “We were getting the full read of 2024 [data] midway through 2025, while we were planning for 2026,” said Vinny Rinaldi, vp of media and marketing technology at Hershey. “That alone is just not conducive to where marketers need to be.”
Irrespective of AI, it's astounding that companies have been throwing $2bn/yr at marketing, and their analytics data on that spend is delayed 6-18 months. That's nearly 15% of their annual revenues spent on guesswork. What's been going on over at Hersheys?
Could be worse. Steve Levitt has a story about working out ad spend causality and actively ignoring the truth[1]:
> LEVITT: And they said to me "[...] One time we hired this summer intern and his job was to do the newspaper inserts for Pittsburgh, and the guy was so incompetent that he just didn’t do it. And when the C.E.O. found that out, he said, ‘If you ever do that again, you’re all fired.’
> LEVITT: So, I said to them, “Well, O.K. But when you looked at the results, what happened to the sales in Pittsburgh when you were dark for a month?” And they called me back about a week later and they said, “You’re not going to believe it. We looked at the data in Pittsburgh, and we saw no impact on sales when they didn’t do any inserts for a month.” I said, “Oh, my God, that’s amazing! O.K., so when can we get started?”
> LEVITT: They said, “Are you crazy?” It was almost if they found out they didn’t work, it was far worse for these people than it was not finding out it didn’t work. Because then they had to explain why for the last 15 years they had been wasting $200 million a year. So, they were happy to just live in a world in which as long as there were ads in every market, every Sunday, life was good.
$2bn is their entire marketing spend, so it includes media buying, creative, in-store displays, PR, research, and anything else involved in the marketing mix.
FTA:
> The confectionery giant, home to brands like Reese’s and Skinny Pop, is working with the analytics platforms Mutinex and Tracer to automate marketing mix modeling — a statistical technique that measures how media spending and other variables drive sales — making it faster and more frequent.
So this system doesn't cost $2bn, it goes into the decision-making of where to spend the $2bn.
I agree that when you have a $2bn budget, it's hard to fathom tolerating a 6-mo analytics lead time, but I'm sure it's not alone, and I'm sure that's why this vendor lobbied Ad Age to cover this project.
My guess is they have to wait to measure the downstream impact of their marketing. And marketing attribution is notoriously fuzzy. If a tv ad runs on may 1 what is the attribution window to someone buying a Hershey bar? How do you know they saw the ad? And that the ad influenced their purchase behavior? The answer is you really don’t do you come up with some kind of formula based a few assumptions.
Ultimately I am highly skeptical. Ad tech is almost always a repackaging of the same product with a new name. They may be using AI to help analyze the data, but I doubt it really has any kind of sizable impact.
What AI actually does
Mutinex has built what it describes as a “multi-agent system,” where each agent acts as a domain specialist. For example, one agent understands marketing econometrics, another understands competitive pricing theory, another diagnoses model failures.
By combining Tracer, which cleans and makes sense of Hershey’s data infrastructure, with Mutinex’s AI system, Hershey is now able run models in as little as three weeks.
In practice, that means faster iteration on how marketing spend is evaluated and adjusted, rather than waiting for lagging historical reads.
“Most companies don’t have an AI problem. They have a data readiness problem,” said Sarah Martinez, chief commercial officer, Tracer.
</article excerpt>
Instead of the headline, it sounds like they've hired an external company to clean up their ETL pipelines. That seems useful.
I'm going to doubt spooling up <massive LLM> with <appropriate system prompt> is going to be the thing that reduces their analysis time.
I am working with agentic AI on industrial manufacturing data. The speed at which you can get insights and dig into all kinds of rabbit holes is insane. It's just as easy to compare plants so you can make strategic decisions on budget allocation as it is to do root cause analysis on why during a given shift there were so many breakdowns.
And this happens with a natural language interface, instead of Excel (although people of course still want an export to Excel button) or worse: by having to go to the BI analyst, have them change a dashboard and after waiting for a few weeks hope they give you what you want...
Yes, you need to structure your data well. Especially metadata/defintions and accessibility - which is not cleaning up ETL pipelines, although that will help. And obviously have lots of relevant data available already (which was my job before this).
From my experience: fully automated marketing budget allocation... Doubt it. Time to insight reduced >10x? For sure.
I know that articles like this, and the broader dialog, doesn't go too in-depth on how these things work. The phrase, "one agent understand marketing econometrics" from the article makes me wonder what exactly they mean.
This could be anything from "we put a prompt in front of chat-gpt that says 'you are an expert in marketing econometrics'" to "we built a model trained exclusively on marketing econometrics material"
No matter what they actually did, the agent (assuming its an LLM) doesn't understand marketing econometrics, instead it's tuned to produce output tokens that I suppose make more sense when the topic is marketing econometrics.
I'm not an LLM detractor, but I find the kind of thinking that's become prevalent to be so squishy. Humans are great anthropomorphizers and it seems today that no one is attempting to hit the brakes on that instinct. The models don't understand anything, in the way that we commonly use the term understand.
It seems we've confused ourselves because the box that doesn't understand marketing econometrics can produce marketing econometrics analysis and when we ask it why it came to such and such conclusion it can produce convincing explanations.
As an aside, I also feel like I've heard this for 30 years about marketing. Marketing is everywhere, the surveillance economy tracks our every move in more and more invasive ways every day, and still companies go "Aw shucks, we just can't make sense of this data." It reminds me of a time when I was working for Abercrombie and Fitch and there was this massive report our team was partially responsible for generating. 500+ pages, generated everyday, sent to a high speed printer from a COBOL job every morning at 5am so that 10-15 copies could be made for the executives. It had *all the data* and each executive had their own little ritual around which bits they thought were important.
Throughout my career as an engineer I've been asked to get more data, more data, more data. Process the data, analyze the data, create some graphs and tables and help people understand the data.
One thing I've realized is that the people demanding all this data, all this insight, all this analysis, are unlikely to actually need it or use it when available. They are tasked with making decisions, and decisions are scary because you can make the wrong one. They would love to not make the decision and maybe you can find enough data that the choices get cut down to just one. Then if it ends up wrong they aren't to blame, the data made them make that decision.
So all of this surveillance, all of this analysis, all of this data is likely just to make some person feel a bit more comfortable about making a decision.
I'm amazed that Hershey is spending $2B on marketing. I honestly don't think I've seen any advertising for the brand itself at any point in the past decade, and I don't have any positive association with it: it's the "bad" chocolate you buy for cooking?
Looking at their subsidiary brands, I don't think I've seen any serious advertising for them either. KitKat is doing viral marketing on social media, but surely, that's not $2B... I've never seen anything for Twizzlers, Pirate's Booty, Milk Duds, Whoppers... the only reason I know these brands is that they're on the shelf in my grocery store.
Unless I'm an outlier, I think it's smart of them to stop throwing $2B down whatever hole they're throwing it into right now.
They would get a lot more ROI by just investing in making products that don’t taste like shit.
Note that some of their products have been enshitified so much that they can’t even legally be called chocolate in some jurisdictions. It’s cheap filler designed to simulate chocolate.
That's specious to state and not at all how marketing works. What if Hershey finds their chocolate is the best thing for specific ailments or an edge case?
"Our Kisses are all you need to send in this pandemic"
All they need to discover is some edge case and take it to the moon. That's how you move shit product.
Well, F just got a boost for an analyst noticing that have a connection to AI based on their batteries and boom.. AI pivot. All of it makes me super nervous about the market.
Zero disagreement on Hershey. It got a lot worse since my childhood days.
Yes, it’s a product of the typical game of gradually swapping out expensive ingredients for less expensive ones and hoping folks don’t notice the gradual change. Over time the product ends up being nothing like what it once was.
This is an example of the same sort of phenomenon as shrinkflation, which they seem to have dubbed "skimpflation"[1] in this article although I've never heard that term used before.
That is it. I heard a possible urban legend that the reason Heshey tastes like that is back in the olden times before the internet they also didn't have refrigerated transport for dairy produce so the milk would go slightly sour when it was transported from the farm to the chocolate factory. This led to the "distinctive"/vomit-like taste of their chocolate and they decided to artificially sour the milk to stick with that flavour even after refrigerated dairy trucks were invented (for some reason).
The “vomit” flavor in Hershey’s base chocolate (which is quite strong for those with good pallets) is butyric acid which comes from aggressive processing of the milk before it’s put in the chocolate. There’s also very little actual chocolate in the final product (compared to other products) which allows these flavors to come out strong vs the chocolate dominating the pallet.
So folks aren’t wrong when they say it tastes like the chocolate is spoiled/off. They’re tasting heavily processed milk.
For people who did not grow up with hershey's, the butyric acid is what makes it taste off.
This is not a question of safety - it is a question of results. They stayed with old technologies and have optimized for cost, not flavor. (And yes, that processing was necessary in the days before a reliable milk source.)
Their products are probably not cherished for any genuine candy-chocolate delicacy across all channels.
By now they should know their place is more "corporate candy bowl" or "look, my date for the night is here" where people about the dopamine pop and fun-chatter.
The truth is that no one can tell why someone bought a Hershey bar. There's no way to distinguish childhood nostalgia from price competitiveness, developed habits, local availability, etc. It's all just guessing*.
The entire ad business lives from this guessing process, so now they can have AI do the guessing.
*Source: I was employed briefly at a company in this space that works with popular brands. They offered me cash to go away when I started sniffing out the underlying BS of how it worked.
People like to claim you can measure advertising and marketing directly and for most products and services that is just not true. But executives think it must be true and so demand it and so they end up hiring / going with the best liar.
Irrespective of AI, it's astounding that companies have been throwing $2bn/yr at marketing, and their analytics data on that spend is delayed 6-18 months. That's nearly 15% of their annual revenues spent on guesswork. What's been going on over at Hersheys?
> LEVITT: And they said to me "[...] One time we hired this summer intern and his job was to do the newspaper inserts for Pittsburgh, and the guy was so incompetent that he just didn’t do it. And when the C.E.O. found that out, he said, ‘If you ever do that again, you’re all fired.’
> LEVITT: So, I said to them, “Well, O.K. But when you looked at the results, what happened to the sales in Pittsburgh when you were dark for a month?” And they called me back about a week later and they said, “You’re not going to believe it. We looked at the data in Pittsburgh, and we saw no impact on sales when they didn’t do any inserts for a month.” I said, “Oh, my God, that’s amazing! O.K., so when can we get started?”
> LEVITT: They said, “Are you crazy?” It was almost if they found out they didn’t work, it was far worse for these people than it was not finding out it didn’t work. Because then they had to explain why for the last 15 years they had been wasting $200 million a year. So, they were happy to just live in a world in which as long as there were ads in every market, every Sunday, life was good.
FTA:
> The confectionery giant, home to brands like Reese’s and Skinny Pop, is working with the analytics platforms Mutinex and Tracer to automate marketing mix modeling — a statistical technique that measures how media spending and other variables drive sales — making it faster and more frequent.
So this system doesn't cost $2bn, it goes into the decision-making of where to spend the $2bn.
I agree that when you have a $2bn budget, it's hard to fathom tolerating a 6-mo analytics lead time, but I'm sure it's not alone, and I'm sure that's why this vendor lobbied Ad Age to cover this project.
Ultimately I am highly skeptical. Ad tech is almost always a repackaging of the same product with a new name. They may be using AI to help analyze the data, but I doubt it really has any kind of sizable impact.
What AI actually does Mutinex has built what it describes as a “multi-agent system,” where each agent acts as a domain specialist. For example, one agent understands marketing econometrics, another understands competitive pricing theory, another diagnoses model failures.
By combining Tracer, which cleans and makes sense of Hershey’s data infrastructure, with Mutinex’s AI system, Hershey is now able run models in as little as three weeks.
In practice, that means faster iteration on how marketing spend is evaluated and adjusted, rather than waiting for lagging historical reads.
“Most companies don’t have an AI problem. They have a data readiness problem,” said Sarah Martinez, chief commercial officer, Tracer.
</article excerpt>
Instead of the headline, it sounds like they've hired an external company to clean up their ETL pipelines. That seems useful.
I'm going to doubt spooling up <massive LLM> with <appropriate system prompt> is going to be the thing that reduces their analysis time.
And this happens with a natural language interface, instead of Excel (although people of course still want an export to Excel button) or worse: by having to go to the BI analyst, have them change a dashboard and after waiting for a few weeks hope they give you what you want...
Yes, you need to structure your data well. Especially metadata/defintions and accessibility - which is not cleaning up ETL pipelines, although that will help. And obviously have lots of relevant data available already (which was my job before this).
From my experience: fully automated marketing budget allocation... Doubt it. Time to insight reduced >10x? For sure.
This could be anything from "we put a prompt in front of chat-gpt that says 'you are an expert in marketing econometrics'" to "we built a model trained exclusively on marketing econometrics material"
No matter what they actually did, the agent (assuming its an LLM) doesn't understand marketing econometrics, instead it's tuned to produce output tokens that I suppose make more sense when the topic is marketing econometrics.
I'm not an LLM detractor, but I find the kind of thinking that's become prevalent to be so squishy. Humans are great anthropomorphizers and it seems today that no one is attempting to hit the brakes on that instinct. The models don't understand anything, in the way that we commonly use the term understand.
It seems we've confused ourselves because the box that doesn't understand marketing econometrics can produce marketing econometrics analysis and when we ask it why it came to such and such conclusion it can produce convincing explanations.
As an aside, I also feel like I've heard this for 30 years about marketing. Marketing is everywhere, the surveillance economy tracks our every move in more and more invasive ways every day, and still companies go "Aw shucks, we just can't make sense of this data." It reminds me of a time when I was working for Abercrombie and Fitch and there was this massive report our team was partially responsible for generating. 500+ pages, generated everyday, sent to a high speed printer from a COBOL job every morning at 5am so that 10-15 copies could be made for the executives. It had *all the data* and each executive had their own little ritual around which bits they thought were important.
Throughout my career as an engineer I've been asked to get more data, more data, more data. Process the data, analyze the data, create some graphs and tables and help people understand the data.
One thing I've realized is that the people demanding all this data, all this insight, all this analysis, are unlikely to actually need it or use it when available. They are tasked with making decisions, and decisions are scary because you can make the wrong one. They would love to not make the decision and maybe you can find enough data that the choices get cut down to just one. Then if it ends up wrong they aren't to blame, the data made them make that decision.
So all of this surveillance, all of this analysis, all of this data is likely just to make some person feel a bit more comfortable about making a decision.
Looking at their subsidiary brands, I don't think I've seen any serious advertising for them either. KitKat is doing viral marketing on social media, but surely, that's not $2B... I've never seen anything for Twizzlers, Pirate's Booty, Milk Duds, Whoppers... the only reason I know these brands is that they're on the shelf in my grocery store.
Unless I'm an outlier, I think it's smart of them to stop throwing $2B down whatever hole they're throwing it into right now.
Note that some of their products have been enshitified so much that they can’t even legally be called chocolate in some jurisdictions. It’s cheap filler designed to simulate chocolate.
Agentic AI is not going to solve that.
"Our Kisses are all you need to send in this pandemic"
All they need to discover is some edge case and take it to the moon. That's how you move shit product.
Zero disagreement on Hershey. It got a lot worse since my childhood days.
https://www.which.co.uk/news/article/shrinkflation-the-brand...
So folks aren’t wrong when they say it tastes like the chocolate is spoiled/off. They’re tasting heavily processed milk.
For people who did not grow up with hershey's, the butyric acid is what makes it taste off.
This is not a question of safety - it is a question of results. They stayed with old technologies and have optimized for cost, not flavor. (And yes, that processing was necessary in the days before a reliable milk source.)
By now they should know their place is more "corporate candy bowl" or "look, my date for the night is here" where people about the dopamine pop and fun-chatter.
The entire ad business lives from this guessing process, so now they can have AI do the guessing.
*Source: I was employed briefly at a company in this space that works with popular brands. They offered me cash to go away when I started sniffing out the underlying BS of how it worked.
It sounds like you got fired and offered severance.
She's right. Companies are too afraid to face the real data from their customers, so they need to hallucinate the data they wanted to imagine instead.
Isn't that what they were doing before AI anyways? It's natural, then.
https://static-www.adweek.com/wp-content/uploads/2026/04/Her...