Subscribe on

Enjoy this episode? Leave us a review.

All Episodes

Episode 167

Synthetic Research and the Future of Marketing Effectiveness with Peter Weinberg

95% of senior marketing and insights leaders plan to use synthetic data within the next 12 months. AI-generated research is already reshaping how brands understand their audiences.

In this episode, Elena, Angela, and Rob sit down with Peter Weinberg, co-founder of Evidenza, to explore how synthetic research is replacing guesswork in B2B marketing, why AI creativity is more formula than magic, and why mental availability still drives purchase decisions in a world of AI recommendations. Peter also shares how Evidenza applied Ehrenberg-Bass principles to its own brand and explains why real-time data obsession can make marketers worse at their jobs.

This video is hosted on YouTube and requires cookie consent to display.

Topics Covered

• [00:47] How synthetic research replaces guesswork in B2B marketing

• [04:00] Accuracy concerns and how purpose-built models outperform general AI

• [09:00] Where to start experimenting with synthetic audiences

• [16:00] Why AI is more creative than most marketers believe

• [22:00] Why brand building matters more, not less, in an AI world

• [27:00] How Peter applied Ehrenberg-Bass principles at Evidenza

• [32:00] Why real-time data can make marketers worse at their jobs

Resources:

The Synthetic Research Breakthrough: How Fine-Tuned Models Outperform General AI (Qualtrics, 2025)

Peter Weinberg on LinkedIn

Today's Hosts

Elena Jasper

CMO

Rob DeMars

Chief Product Architect

Angela Voss

Chief Executive Officer

Peter Weinberg

Co-Founder of Evidenza

Transcript

Elena: Hello and welcome to "The Marketing Architects," a research - first podcast dedicated to answering your toughest marketing questions. I'm Elena Jasper. I'm on the marketing team here at Marketing Architects, and I'm joined by my co - hosts, Angela Voss, the CEO of Marketing Architects, and Rob DeMars, the Chief Product Architect at Misfits & Machines.

Angela: Hello.

Elena: And we have a guest joining us today, Peter Weinberg, co - founder of Evidenza, an AI market research company leading the synthetic revolution. Peter was one of the main minds behind the LinkedIn B2B Institute's research arm, which produced some of the most influential market effectiveness work of the last decade. Peter, welcome back to the show.

Peter: Delighted to be here. Thank you guys so much for having me.

Rob: Okay, I learned something fascinating in my prep. You call yourself a profoundly unsuccessful science fiction novelist, which I love. I love that because your day job is literally building synthetic humans inside a computer, and your side project is writing about synthetic humans inside a computer. So when do we get to see the book?

Peter: Yes, I'm very proud of how profoundly unsuccessful I was. I spent nine years writing a science fiction book. I have absolutely nothing to show for myself other than a finished copy. You know, Amazon will actually print the book for you, so I have a printed version. The problem — nobody read it.

Rob: What's the title?

Peter: It's called The Fall of Newest York. It's like an apocalyptic, futuristic Manhattan. I would say my business is kind of a science fiction-y type business, and that has been successful. So I would say non-fiction science fiction has been a good path for me. Fiction science fiction, less promising, I'm sad to say.

Angela: Well, I think you're winning.

Rob: I would say so, yes.

Elena: Yeah. I also think actually sitting down and writing a book is very impressive because I think a lot of people say they want to. So even sitting down and finishing one...

Peter: You know, that's kind of how I feel about it. It's like, you gotta just minimize your regrets in life. I would've regretted my whole life if I had never attempted to write a novel, 'cause that was always my dream — to be a great novelist — and now I won't have that regret, nor will I be a great novelist, but at least I tried. At least I tried.

Rob: Well, Angela and Elena both read, so you should send them a copy, and then I'll make the audiobook version of it. You guys could be the second, third, and fourth people to have ever read the book other than my mother, so that would be amazing. I mean, that'd be a 4X increase in sales. That'd be super exciting.

Elena: Yeah, send me a copy. I'll check it out. All right, we're back with our thoughts on some recent marketing news, always trying to root our opinions in data, research, and what drives business results. Today, we're talking about marketing effectiveness broadly, but also synthetic research — the idea that AI can simulate consumer audiences and produce marketing intelligence faster and cheaper than traditional methods — and what that means for the future of how brands make decisions.

But first, I wanna kick us off, as I always do, with some research. Today's episode was inspired by a piece I read from Qualtrics titled "The Synthetic Research Breakthrough: How Fine-Tuned Models Outperform General AI." This was published a couple months ago, and the article reported that 95% of senior marketing and insights leaders say they're already using or plan to use synthetic data within the next 12 months to do things like generate new insights, fill data gaps, and replace or augment traditional surveys.

The top reasons they're doing this are speed to insights and better depth of insights. The article also acknowledges some skepticism behind these tools. Early attempts to use general-purpose LLMs for research seem to be too agreeable. They might have lacked some demographic variation and failed to capture real human nuance.

But the Qualtrics team found that purpose-built, fine-tuned models trained on domain-specific data produce results nearly identical to human responses out of the box and significantly outperform general-use LLMs. So Peter, thanks again for joining us. You left LinkedIn. You've built this synthetic research company, Evidenza. To start, I'm curious — how are your customers using AI and synthetic audiences to improve or sometimes even replace their traditional market research processes?

Peter: Yeah, that's a great question. I mean, well, I would say ideally when you make a marketing decision, it is informed by the voice of the customer, right? That's not a controversial idea. That's the famous Jeff Bezos idea of the empty chair — that you're supposed to leave an empty chair in every room, and it's supposed to represent the customer, and you're supposed to kind of look at the chair and ask yourself, "What would our customer want us to do?"

Right? Because the customer decides if the marketing works or if the product is bought or if the sales pitch is successful, right? The customer determines everything. But there's a reason the chair is empty, which is that it's deeply impractical to talk to a customer every time you need to make a decision.

And that is doubly, triply, quintuply true if you're a B2B marketer. So if your core customer is, let's say, a CTO of a large financial services company, trying to get those people to take market research surveys in exchange for, like, a $50 Applebee's gift card is very, very hard. Dare I even say impossible, 'cause these tend to be, first of all, very small audiences and very highly paid professionals who are very busy, and they can buy their own meal at Applebee's should they want to, right?

So, you know, as a result, most B2B marketers actually don't do market research, especially quantitative market research, because they cannot get a sample together. It's too slow. It's too expensive. So I think the interesting thing about synthetic research and what we do — you know, AI-generated research — on the one hand, you could say it's replacing traditional research, and I'm sure there is some truth to that.

But that's actually kind of not the core use case in my mind. What it's actually replacing is, like, vibes or ignorance. In other words, most of these marketing decisions that were being made, especially in B2B, they weren't really based on any research because people just don't do quantitative market research. It's too slow. It's too expensive, right? So I think what you're actually seeing is, for the first time, you can make a marketing decision not based on what did the sales team tell you or what did your mom tell you, but what is the customer — in this case, the synthetic customer — telling you. And that, I think, is a very big disruption.

Elena: Yeah, that's a good point that for a lot of marketers, they're not using any sort of traditional research processes. It's just giving them information they wouldn't have had before. With anything new, there's always skepticism. Especially with AI, some people hear anything associated with AI and they automatically have some negative emotions towards it. What do you think have been the biggest misconceptions or ways that marketers still underestimate synthetic research today? 'Cause it's been, you know — you've been around for years now — but it still feels like that's out there. So where do you think those biggest pockets are?

Peter: Yeah, I mean, listen, it's a new category. In any new category, there will be early adopters and late adopters and everywhere in between. I think as it relates to synthetic research, nobody needs to be convinced it's faster and cheaper. People believe that right away. But there is definitely a lot of skepticism around accuracy — and, you know, will a robot CFO of a large financial services brand give you the same answers as flesh-and-blood CFOs, right?

So I think fundamentally it's about accuracy. You know, I think it's a reasonable concern, right? Like, I think you should have some skepticism towards it. There's just a healthy skepticism where if you see evidence to the contrary, you can change your mind, and then unhealthy skepticism where you just refuse to believe it's even possible because it's such a science fiction-y type idea.

I think the way we think about accuracy is this: so what we basically do is we look at what humans say in a quantitative survey, and then we look at what synthetic respondents say in a quantitative survey, and we compare them. We look at basically the similarity in terms of the decisions that are being made. What's the top choice and the bottom choice? We look at correlations, we look at mean difference, we look at a range of different statistical measures. And we've now done that hundreds of times across hundreds of categories and markets.

And what we find pretty much every time is that the synthetic answers very closely track the human answers. So I think it is reasonable to be concerned around accuracy, and I think that's the biggest concern there is in synthetic research. I just think there are ways to verify the accuracy through testing.

And I think, you know, to the research you brought up at the beginning of the call, just 'cause somebody tried to tinker around in ChatGPT by themselves and couldn't get synthetic research to work, that doesn't mean synthetic research doesn't work. It means they couldn't get it to work, or it means that general all-purpose AI, like a ChatGPT tool, can't necessarily do it with statistical reliability.

But if you build a specialized system that is custom-designed to do synthetic quantitative research for B2B — which is what we've built at Evidenza — then I think you can get to very reliable outputs that help you make decisions faster and much more cost-effectively than you could before.

Elena: Yeah, it seems like a really important piece of building something like this is having data to validate, but then once you have it, why would you not scale this and use it as often as you can? If a marketer's listening to this and they're like, "Oh, I'm excited. I wanna test some synthetic audiences," — obviously they should work with Evidenza first — but if they were just trying to get started, where could they begin, and what would be some watch-outs as you're starting to experiment with synthetic audiences?

Peter: Yeah. Well, thank you for the advertisement — the earned media. We hugely appreciate it. Yeah, I think I go back to the top where you're talking about traditional market research. I think if you already have market research on a topic and you have a good human market research solution there, I don't think that's where you should start.

I think you should say, "What's a decision we need to make where we really have no good data?" Like, it's a new audience we don't understand, it's a new product we're trying to launch, and we're kind of flying blind. And that's where I think you can now bring synthetic market research to bear. So it's kind of like — what are the decisions where there's a gap or a lack of research? And that's where I would start rather than, you know, "Hey, I'm already doing this brand tracker every year with human research. Can I do that synthetically?" That's a tougher transition because you already have a solution in place. I think the white spaces in my mind are usually the best places to start, and there's a lot of white spaces in B2B because as I was saying, I think something like 90% of the time the marketing decision was not based on any market research to begin with.

Elena: Yeah. It's good advice too because then you're not taking someone's baby or something.

Peter: Yeah, exactly. I mean, that's exactly right. It's like a political thing, right? It's like some people have a very strong vested interest in the traditional market research they're already doing. You know, maybe try to find a place where nobody has a vested interest, and that, I think, is where you find more openness to new approaches, shall we call it.

Elena: So I'd say — you mentioned this is a new category. Marketers definitely, and the people that are using this, I think are kind of the earlier adopters. It's gonna continue to become popular. I know this can be an annoying question, but where do you see this going, like, the next couple years? Do you think we're gonna be able to do new things with synthetic research? Do you think this is gonna become commonplace? What do you see — when you're looking towards the future — what's next for synthetic audiences?

Peter: I don't think it's an annoying question. I think it's a great question. As a lover of science fiction, I love thinking about the future. You know, I think it's funny. This idea of synthetic research, or AI-generated research, it sounds very futuristic, but it's actually not. Synthetic substitutes is a very, very old idea, and they have transformed dozens of industries over the past hundred years. So, like, to give you an example, back in the day, vanilla was like a luxury spice for kings and queens because it grew in equatorial orchids or something like that. Then in, like, the 1890s, they figured out how to make synthetic vanilla. Now everyone has vanilla in their kitchen cabinet, right? Rubber is a fun example. You know, the Allies needed to scale production of rubber in World War II. Growing rubber on trees was not a super scalable way to do that, so somebody figured out how to make synthetic rubber. Diamonds is a famous example, right? Diamonds used to be grown by Mother Earth over millions of years — not particularly scalable, very, very expensive — until General Electric in, like, the 1950s figured out how to make lab-grown diamonds.

And now, actually, I think last year was the first year ever where sales of lab-grown diamonds eclipsed sales of natural diamonds, and the complaint everyone makes is that they're too big and too cheap and too perfect, right? Insulin may be a feel-good example. Like, insulin used to come from the pancreas of pigs, and then they figured out how to make synthetic insulin and sort of scale insulin to millions of diabetics all over the world.

So there is kind of just a pattern here that I see whenever a new synthetic substitute comes, which is — first of all — it actually tends to start in B2B categories, which is interesting. So, like, the early applications of diamonds, it wasn't for engagement rings. It was GE wanting it for mining equipment because diamonds are, you know, the hardest objects.

So it often starts in B2B, over time scales into consumer, and over time it increases the accessibility and size of the category because the cost comes down, the speed comes up, and over time things just become more and more synthetic. So that is exactly what I think you will see play out in synthetic research.

I think it will, and has, started in B2B. I think it will scale to B2C and every market and every category. And ultimately, I would expect a future where, like, let's say 90% of market research is synthetic market research, which of course is not true today. But again, I think if you just play out what happens when you introduce a new version of a category that is machine-made instead of man-made, that is typically what happens, and that is what I would expect to see happen here.

Rob: Those were some really interesting examples. I don't know about you guys, but I still prefer my insulin to come from a swine.

Peter: I agree. There's no insulin like pig pancreas insulin. I've been saying that for years. Well, that's why — you know — there will always be people who just want the pig pancreas insulin.

Rob: I want the pig insulin. Yeah, exactly.

Peter: I want the pig insulin — always gonna be a market for it — but I think it's gonna be a shrinking market if I had to guess.

Rob: I'd agree. And, you know, when it comes to synthetic research, it does seem like the pendulum's starting to now swing towards it becoming a much more accepted form and use of AI. But there are still other areas that I think are still controversial for some folks, like being able to say that AI just can't be creative, right?

And I've heard you talk about on your own show that you would disagree with the idea that AI can't be creative. Will you talk more about that?

Peter: Yeah, absolutely. I mean, I think there are so many misconceptions around AI. I often like to think of them as kind of like coping mechanisms. Like, ideas that are scary — people are like, "Oh, you know, I don't want AI to be creative, so AI isn't creative." Right? Is that cope or is that hope? Right? I do think AI can be creative because I think creativity is not this magical lightning-in-a-bottle thing that everybody seems to think it is. You know, people are like, "Creativity — there's no science to it, there's no formula to it, you know, could never be done by a machine."

Rob: And then I always like to think about Disney movies. Big fan of Disney movies, as are my very young children. So, like, let's talk about WALL-E, Finding Nemo, and The Lion King. You'd be like, "Oh my God, those movies are so creative. I can't believe how creative they are." And then you kind of look under the hood and you're like, "What's The Lion King about?" It's about a lion that gets lost in the jungle and needs to find its way home. What is Finding Nemo about? It's about a fish that gets lost in the sea and needs to find its way home. What is WALL-E about? You may recall it's about a robot that gets lost in space and needs to find its way home. All vaguely reminiscent of this book about a Greek guy who gets lost and needs to find his way home. Not sure if you've heard of that one — The Odyssey. A book that nobody read, but I'm sure they'll see the movie. Christopher Nolan, you know it.

Peter: Right. So all of that is just to say, like, there are patterns to effective creative. There are formulas in terms of — structurally — what makes for good creative. If you think about marketing effectiveness, there's this idea of brand codes or distinctive brand assets, which are basically creative patterns that you repeat in an execution to make sure you get credit for the ad and not the competitor.

So all of that is just to say I think these things are more formulaic than creatives would have you believe. And, you know, I see creative directors being like, "Well, it's just trained on the past, so it can't be creative." But that's really all creativity is — it's the recombination of things from the past.

That's like the famous Picasso idea that genius steals, right? Or like Mark Twain said, "There are no new ideas." You just put old ideas through a kaleidoscope, essentially.

So if you just think about creativity as recombination of things that happened in the past and sort of applying formulas, I think AI is going to be very good at that because, you know, you haven't read every novel ever written, but AI has. You haven't read every white paper that's ever been written — AI has. You haven't seen every ad that's ever been aired — AI has.

So it actually has a much broader pool to recombine against. And I think there are more and more examples of AI being creative. You know, going back to AlphaGo, where the robot made a move in the game of Go that the human player had never thought to make, and ended up being the winning move — move 37.

You've got new medications being developed by AI. So the idea that AI can win at Go and it can cure cancer, but it's not going to be able to come up with advertisements? I would classify that as a coping mechanism personally, as unpopular an opinion as I'm sure that is.

Rob: I mean, it can fold protein for crying out loud.

Peter: Right. It'll fold protein, but a 30-second TV spot? No chance. No chance the genius robot will be able to figure that out.

Rob: We do think we're so special, and we've all been in the brainstorming meeting where someone throws out an idea and you're like, "That's an amazing idea — when Nike did it three years ago." You know? So it's like we all borrow from frameworks and the hero's journey, to your point, and...

Peter: We do. Also, like, who kills the great ideas in most marketing departments? The marketers kill them, right? 'Cause they're too creative. The interesting thing about AI is that it's not self-conscious, and in many ways it's willing to consider crazy ideas that the human wouldn't consider, right?

Rob: For sure. Well, moving on to things like video generation and this sort of phenomenon we call AI slop — do you think that's overblown? That's a term that, you know, another defensive term that people will throw at marketers who are making TV commercials using AI.

Peter: Yeah, hugely overblown. I mean, again, you're asking me some of my least popular opinions, but I'm always happy to share them. I just think it's very funny. I think you have these marketers being like, "Oh no, we're gonna be — all the ads are gonna be slop, and they're all gonna be so bad and terrible, and we're not gonna have great advertising anymore."

Like, just to be clear, that's the world we currently live in. Like, the vast majority of advertising is slop. It's just produced by humans at great cost and over a long period of time. There is quantitative, empirical research on this. Like, when I was at the B2B Institute, we measured 5,000 B2B ads. We looked at them on a scale from one to five — five being very effective, one being very ineffective. 80% of them scored one on a one-to-five scale, right? So, like, the ads are already bad. They're just being produced by humans. So I think, you know, if you want a really cynical take, it's like, okay, well, maybe it's not gonna get any better, but at least it's gonna be cheaper and faster.

If you're gonna have slop, I'd rather have AI slop than human slop if I were a CMO trying to make my marketing budget as efficient as possible. But again, I don't think it's going to be slop. I think it's going to be better. You know, I think there's this key idea of temperature in AI that people are not very familiar with. Have you come across this concept of temperature?

Rob: Yeah, in the model — if you're referencing the same thing — we have a pre-testing platform called ScriptSooth. Rob was instrumental in the build, and that's where we played with temperature quite a bit.

Peter: That's a great application of it. So temperature is like — you could think of it as a creativity dial for AI, right? Where if you turn the temperature way up, you get weird outputs, and if you turn it way down, you get very tame outputs. So if you say, you know, "Tell me a bedtime story," — if the temperature is set very low, it's gonna say, "Once upon a time, there was a princess in a castle," right? Very not creative. If you turn it up to a 10, all of a sudden it's like, "One time Cat Stevens flew an avocado to Mars to rescue a mushroom," or something like that, right?

Rob: How much weed do you want the model to smoke?

Peter: Exactly. How much weed do you want the model to smoke? And I think what people don't understand is that on the generic LLMs — like on a ChatGPT — the temperature is preset usually by the model, and it's set very low because people will get upset if ChatGPT starts to give weird answers. They'll say it's hallucinating, and they'll be upset about it, so they tamp it way down. But to your point, like, there are some cases where you want the temperature way up. You know, if you're trying to come up with new ideas for an ad or for a product, all of a sudden you might wanna turn the temperature up to 10, which you can do if you're working with the APIs and the code of an LLM instead of these sort of off-the-shelf tools.

So all of that is just to say, like, you know, some AI may not be creative by design, but there are things you can do to make AI more creative. And in many cases, again, I think it's going to be more creative than a human creative could ever be.

Rob: Well, speaking of all of this as it relates to the development of a thing that we all hold true to be an important brand — right? Brand is still central to all that we're doing with marketing — and you've talked a lot about that. Why do you believe brand will continue to remain important even with all this incredible advancement we have going on in marketing?

Peter: Yeah, it's a great question. You know, again, when you think about the future, you should think about the past. The reason brands are successful, the reason brands help companies sell products, is due to a fundamental limitation of the human brain, which is that people are lazy, and they don't like to spend a lot of time thinking about their decisions, so they just kind of default to things that are mentally available and in their head.

So I'm thirsty, I need a drink, you know — I think of Coca-Cola. It just comes easily to mind, right? The human brain hasn't changed in hundreds of thousands of years. The invention of AI is not going to change the way the human brain works fundamentally. The hardware is not gonna get an upgrade.

So you just have to imagine — no matter how advanced the technology becomes — assuming you still have a human brain making the decision about what brand to buy, whether it's B2B or B2C, it's gonna default to lazy choices, and that's what a brand is. It's a lazy choice. So I always think about it like this.

It's like the most valuable search engine is the one in your head. Somebody said that once, right? That's the most valuable search engine — the one inside your brain. When you're in a pub in Ireland and you need a beer, you don't open up Google and say, "What beer should I get at this Irish pub?" And you don't go to ChatGPT either. You know, you just search your brain, and it turns up Guinness, and you order an absolutely delicious Guinness. So good. So it's a lovely day for a Guinness today, by the way. Every day is a lovely day for a Guinness.

So the point is, like, yes, I'm sure people will search in LLMs more and more, and I'm sure that will eat traditional search more and more. But at the end of the day, that's not where the decision starts. The decision starts in your brain, and the brain is searching for brands. Even in a world where LLMs are surfacing brands, you're also going to be much more likely to accept the suggestion from the LLM if it's a brand you've heard of. So if I go into ChatGPT and I say, "Hey, I need an IT service management software for my company," and it says, "Yeah, I've got two good options for you. I've got ServiceNow, and I've got Jimmy's Discount ITSM Shop — which one would you like to buy?" Right? Like, it could even say, "Hey, I swear by Jimmy's. Jimmy's is really good." But then I'm like, "Huh, well, I've never heard of Jimmy's," and when I go back to my CTO and suggest Jimmy's Discount ITSM Shop, not sure he or she is gonna go for that choice, so I'm gonna go with ServiceNow, right?

So my point is you will think of the brand first. You will buy the brand you think of. Even when the LLM suggests a brand, you will default to the brand that you're already familiar with. And for all those reasons, I think brand is an immortal idea. You know, it's been important for thousands of years, and it will continue to be important for thousands of years because humans are hardwired to prefer familiar brands.

Angela: Well, that's the reason — one of the reasons — that I get so excited about AI, because as marketers, we have constraints that are put upon us. Could be budget, could be time. We think about how we create mental availability and what are the category entry points that we maybe don't get to — they never work their way into the ad campaign. And how do we use AI? Like, it's such a fun marketing challenge. That's why it's so frustrating to hear all the skepticism around it, but preaching to the choir.

Peter: Well, I think to your point, Angela, AI will help you build brands better than before because, for example, with synthetic research, you can now identify what are the category entry points for your brand, what is your mental availability, what is the account-based segmentation, and what accounts should you target with category entry points?

So I think brand was always important — that's not changing — but the efficiency and effectiveness with which you can build brands I think will be much higher because of AI, not lower.

Angela: 100% agree with you there. You've spent years championing these principles that actually make B2B marketing work — all marketing really — but you've kind of leaned into the B2B side. A lot of that groundwork was laid during your time at LinkedIn with John. So I'm curious — now that you have built and stood up Evidenza from scratch, how did you look to apply some of those same principles to your own brand? You know, where do you start when you're kind of the one in the hot seat there? It's a fun challenge.

Peter: Yeah, it is funny. You know, I feel like I spent a lot of my career advising marketers and trying to get them to apply Ehrenberg-Bass principles, and now I am the marketer, right? So I need to decide how to apply it for ourselves. And I would attribute a lot of our success to us applying Ehrenberg-Bass principles.

I mean, the core principles are mental and physical availability. So are you easy to mind and easy to find? Like, those are the two pillars of marketing effectiveness, and we've worked on both. You know, when I think about mental availability, we have focused on reach. So I'm trying to reach as many category buyers as we possibly can, which in our case is B2B marketers, and we track and optimize for reach religiously.

That's kind of on the distribution front. Then when I think about our creative, we have to speak to the right category entry points, and we do. People are not buying synthetic market research. They don't buy traditional market research either. They have a problem, and that is occasionally the solution, right?

So for us, it's like — what are the category entry points? Hey, I'm looking to launch a new product. I'm looking to enter a new market. I'm looking to go after a new type of buyer. Those are all buying situations or category entry points that trigger you to be in market for market research, and that's where we try to make sure we come to mind. So our marketing is trying to speak to those purchase occasions.

And then I would say probably the place we have applied it the most is in distinctiveness, which is this other key Ehrenberg-Bass concept, which is — you know — you wanna look like yourself and no one else. You don't wanna blend in. You wanna stand out, and that has been core to our company.

So, like, when we first had to come up with the name and the brand, we were like, "Okay, you know, what's something that's going to be weird and grab people's attention?" Well, you don't meet a lot of Italian AI companies. You don't. You know, you meet a lot of Silicon Valley tech companies that are blue and have, like, this cute little flat animation style. You don't meet a lot of Italian-branded AI companies whose founders dress up in futuristic Italian Renaissance costumes on their podcast, who start every email with buongiorno, who finish every presentation with grazie mille. Like, we have leaned really hard into futuristic Italian Renaissance, because it is distinctive to the category.

You don't see anybody else doing it in AI or probably in tech in general. It's also something — I think I've been amazed — when you choose something like that, how consistently you can apply it. So, like, if Italian is our distinctive asset, well, when we wanna host a client event, guess what type of restaurant we do the event at?

Angela: Not going for cheeseburgers, huh?

Peter: Not going for cheeseburgers. We're going upscale Italian, right? You know, if we wanna do a big fancy event somewhere, what country will it be in? Of course, it will have to be probably in Sicily, you know, maybe in Sardinia — who can say? But you know, it's actually been very interesting.

Part of the reason we chose futuristic Italian Renaissance is that it's something the LLMs can understand and reliably replicate. So in other words, you know, we use a lot of AI-generated imagery in our product, in our advertising, in our presentations. And, you know, a lot of brand guidelines that companies have are so vague that they're not really AI-friendly. Like, everyone's like, "Oh, we're bold, modern, and we're sexy," you know? Give that prompt to an LLM, and you're gonna get wildly inconsistent results. Go tell the LLM you want an image that is designed in a futuristic Italian Renaissance style. The LLMs have been trained on every Renaissance painting ever painted and the entire genre of futurism, and it will have no trouble giving you a very consistently distinctive output.

So I would say on the mental availability front, yes — we focus on reach, category entry points, and distinctive brand assets to grow our brand. And then physical availability is just, you know, making sure we're easy to work with, making sure you can easily get in touch with our sales teams, making sure our sales teams respond quickly, reducing all the friction we can in the buying process for procurement — all of that stuff.

So I'm proud to say that I think we have become a minor case study in Ehrenberg-Bass marketing. I think that's what grows big brands. It's what grows small brands, tech brands, finance brands, CPG brands — any brand.

Angela: Absolutely. I know we were talking before we started recording here, but this is Peter's second time on the show, and the first time you were in stealth mode. It was right in between the LinkedIn journey and the Evidenza journey. So it's just been so fun to watch you and John do your thing and see it come to life, and clearly it's working well for you.

Peter: Thank you so much.

Angela: Yes, of course.

Peter: Grazie mille, I should say. Stick to the brand. Sorry, sorry. The brand police are gonna come arrest me. I mean grazie mille. Grazie mille.

Angela: You guys have never been afraid to be the ones in the room saying the uncomfortable thing — contrarians, right? And one of the arguments that has stuck with me is that more data, and more real-time data specifically, isn't just unhelpful — it can really make marketers worse at their jobs in an industry completely hooked on precision. What do you think people are fundamentally getting wrong about the relationship between data confidence and actual good decisions for the brand?

Peter: Yeah. I mean, I still feel that way. It's been an interesting situation for us 'cause we're a synthetic research and synthetic data company, and so we're now able to provide clients with more and more data than they ever had before. But I still, at my core, believe that more data doesn't always lead to better decisions.

So it's like with synthetic research, for example — well, you could now ask in a 1,000-question survey instead of limiting yourself to a 10-question survey. But that doesn't mean a 1,000-question survey is gonna be better than a good 10-question survey, right? In fact, you may get a lot more noise because there's this idea of signal-to-noise ratio, and in any dataset there is signal — things you need to pay attention to — and then noise, which is basically information you need to ignore. And the bigger the dataset, the more noise there is, right?

So I think marketers often react to noise. I think there's a good parallel with the financial markets here, right? Like, if you ever talk to a financial advisor, they will tell you — don't try to time the market, buying and selling stocks at particular times. Instead, focus on time in market — so just buying in all the time and holding over long periods of time. Because if you look at the long term, right, the stocks always go up. The stock market has always gone up. If you look at it day-to-day, you see these crazy jagged lines — it's up, it's down, it's all over the place.

So if you're checking the stock market every day, you're seeing a lot of noise. If you're checking it every decade, you're seeing a lot of signal. And I think the same thing is true in marketing. Like, let's talk about brand tracking. If you ask clients, "How often do you want brand tracking? In your ideal world, how often would you get data on the awareness of your brand?" The answer is like, "Well, ideally minute by minute. I would settle for, you know, hour by hour. Day by day would be cool." No one's gonna opt in for annual brand tracking or quarterly brand tracking. But you are much more likely to see signal in an annual brand tracker than in a daily brand tracker, right?

Because you're more likely to — well — the creative is more likely to have sunk into the market. You're more likely to get a big sample. It takes a while for these things to compound. You know, even at the bottom of the funnel, the same thing is true with metrics like CPL, right? Like, I would have this at LinkedIn all the time where people are like, "Oh, you know, our CPL is too high. We gotta cut the budget." And I'm like, "Cool. How long has the campaign been running?" "It's been about three hours." "How big is the spend?" "About, like, $60," right? So that is the law of small numbers. That is unreliable data. Like, the longer the time horizon on the data and the bigger the dataset, the more reliable it becomes.

So I think the truth is, this obsession with real-time data all the time — you know, what you actually need is all-time data. You want data over long periods of time, much more likely to help you make good decisions than sort of knee-jerk reactions to noise, right? It's like you can be Warren Buffett or you can be a day trader, and I think most marketers are day traders, and day traders have severely underperformed Warren Buffett over the past 100 years.

Angela: Ain't it the truth.

Elena: Thank you, Peter. It was so hard to choose which topics we wanted to cover with you today 'cause I love so many of them on your podcast. But to wrap us up — we talked about kind of the future and science fiction — so final question for you: If AI could perfectly predict the future, what's one thing you'd want to know, and what's something you definitely would not want to know?

Peter: I think it's a famous Niels Bohr quote: "It's very hard to predict things, especially the future." I'm more focused on the past than the future. The marketing industry has too many futurists and not enough historians. That's what I would say. So I'd be more interested in what's happened in the past 100 years than trying to predict what will happen in the next 100 years. So I would just say I'm broadly uninterested in AI's prediction of the future and wouldn't wanna know, you know? 'Cause if you know the future — you've seen enough science fiction movies, I'm sure — once you know the future, then all of a sudden you start doing all kinds of weird things and you disrupt the timeline. You end up in some dystopian timeline. So I'd like to leave the future as a blank slate. Focus on the past. That's my evasive answer to your great question, Elena.

Elena: No, yeah, it's a scary question. Ang, Rob, would you wanna know?

Angela: I don't know. I was thinking like, what would I really wanna know? I think if there was a moment in my life where I knew this decision... You know, sometimes you come across decisions that feel really light and easy. Some of them are really heavy and hard, and you tend to put emphasis on those heavy and hard ones, and yet maybe some of the light and easy ones actually are defining moments. So, you know, what are those key decisions that might have the biggest positive impact on your life? It might just be something like, you know what? In the morning, I'm not gonna do sugar. I'm just not gonna do it anymore. And it has like this profound change on your health — or something like that would be cool to know.

Peter: Yeah, and the butterfly effect, right? The little thing that has a big impact.

Angela: Yep. Conversely, I would say maybe what I wouldn't wanna know is one of the beliefs that I had that I would look back upon with embarrassment or something. It's hard. It's really hard. I'm with you that, like, just living in the now might be the best.

Rob: I've got a really important one for me. I take bladder management very seriously, so if I could predict when I'm gonna have to pee — if AI could predict that for me, I would really value it. And I'm desperate, like I've been trying to run longer, and when I'm running, I need to know. I need to know when I'm gonna be able to dispense of, of things.

Peter: That would be a feature. It's weird none of the AI companies, none of the frontier models, have really prioritized that use case, but it seems like there could be a huge market for it. New business idea for you, Rob.

Rob: The Apple Watch Ultra, you know, there you go. Who's the new guy? It's not Tim Cook anymore. But there's the big feature on the new watch. So definitely could see some AI advancement there. I think the counter side is I really hate movie spoilers. I would hate to know what's gonna happen with Dr. Doom in the next Avengers, so I don't want any spoilers.

Angela: All right, that's fair.

Elena: Yeah. I'm with Peter. I don't really wanna know anything. I was trying to think of something. I'm like, I don't want things spoiled. But one thing I'd love to know — not specifics, but in general — will a Minnesota sports team win a title or championship in my lifetime? And don't tell me who, but if I could just hear that someone was going to.

Angela: But if you found out no — the AI just goes no, period.

Elena: I'm already a Packers fan, so I might just shift my allegiances fully to Wisconsin maybe, or the other sport group.

Peter: Well, you could make a killing in the prediction market. So yeah, it's another product idea.

Elena: There we go. Amazing. Well, Peter, thanks so much for joining us again. That was so fun. We always love talking to you and hearing your insights, so thanks for coming back on the show.

Peter: Thanks so much for having me. Absolute pleasure. Thank you. Thank you. Grazie mille. Grazie mille. And arrivederci.