Retailgentic | Consumer Behavior & Retail Trend

This week’s episode is a deep dive into personalization, a huge category in e-commerce that’s getting rebuilt from the ground up for the AI era. 

FERMÀT is an AI-native personalization platform that sits on top of your e-commerce platform and customizes the site experience based on a visitor’s behavior, not their identity. That distinction sounds subtle, but it turns out to be the whole ballgame and it connects directly to agentic commerce because agents show up with intent, not an identity. We also pick up Rishabh’s founder journey and some very candid company-building advice along the way.

Highlights 
  • 𝗧𝗵𝗲 𝟮𝟬% 𝗰𝗼𝗻𝘃𝗲𝗿𝘀𝗶𝗼𝗻 𝗹𝗶𝗳𝘁: FERMÀT's data shows personalizing on behavior alone, even for fully incognito visitors, converts roughly 20% better than personalizing on identity and past purchases.
  • 𝗠𝗮𝗻𝘆 𝘁𝗼 𝗼𝗻𝗲, 𝗻𝗼𝘁 𝗼𝗻𝗲 𝘁𝗼 𝗼𝗻𝗲: the marketer's dream of one-to-one was "a lack of imagination", the same person is a different shopper on Instagram, Google, and ChatGPT. 
  • 𝗗𝗲𝘁𝗲𝗰𝘁, 𝗿𝗲𝗰𝗼𝗺𝗺𝗲𝗻𝗱, 𝗴𝗲𝗻𝗲𝗿𝗮𝘁𝗲: instead of humans watching a sample of session recordings and shipping site changes months later, agents trained on commerce watch every session, recommend changes, and generate the new experience.
  • 𝗧𝗵𝗲 𝗕𝗹𝗼𝗼𝗺𝗿𝗲𝗮𝗰𝗵 𝗽𝗿𝗼𝘃𝗼𝗰𝗮𝘁𝗶𝗼𝗻: identity-based personalization was state-of-the-art 10 years ago. Rishabh's take on the incumbents: if nothing changes, "their 3 to 400 million of revenue will become 2 to 300."
  • 𝗠𝗶𝗻𝗲 𝘁𝗵𝗲 𝗮𝗱, 𝗻𝗼𝘁 𝘁𝗵𝗲 𝗜𝗗: even when an Instagram visitor is anonymous, the ad they clicked carries demographics, performance history, and frequency data… context most brands never use.
  • 𝗧𝗵𝗲 𝗿𝗲𝗮𝗹 𝗲𝘀𝘁𝗮𝘁𝗲 𝗮𝗴𝗲𝗻𝘁 𝘁𝗲𝘀𝘁: an agent is only as good as the feedback, context, and iteration speed you give it and every "actually, I also care about the collar" update is behavior data.
  • 𝗦𝗮𝘆 𝘄𝗵𝗮𝘁'𝘀 𝘁𝗿𝘂𝗲: Platforms are using AI to verify your pages and truth gets rewarded, not dinged.
  • 𝗦𝘂𝗽𝗽𝗼𝗿𝘁 𝗮𝗴𝗲𝗻𝘁 ≠ 𝘀𝗵𝗼𝗽𝗽𝗲𝗿 𝗮𝗴𝗲𝗻𝘁: support agents should stick to facts and policies; shopper agents should infer.
Rishabh is one of the clearest thinkers I’ve talked to on where personalization goes in an agent-first world, and his founder advice at the end is worth the listen on its own. Enjoy the conversation.

Timestamps: 

02:44 Welcome Rishabh 
03:37 Rishabh's backstory: physics, Wall Street, PhD
06:33 Building three businesses at LiveRamp
07:56 Founding story: Apple ATT and iOS 14.5
10:33 Detect, recommend, generate
12:27 Where FERMÀT sits in the stack
13:57 Many-to-one personalization
15:22 The Bloomreach question
19:00 Does current behavior beat past behavior?
21:18 The data: a 20% conversion lift
22:34 The agentic commerce connection
23:44 The real estate agent analogy
25:41 Reading intent from incognito traffic
26:28 Into the weeds: mining a Meta ad for context
30:37 Side quest: the SEO cloaking myth
33:47 Writing for agents, bots, and humans
34:49 On-site shopper agents
39:40 Company building: know your why
42:38 Where to find Rishabh + the 1 to 100 podcast


👉 Connect with Rishabh: https://www.linkedin.com/in/rishabhmjain/
👉 Learn more about FERMÀT: https://www.fermatcommerce.com/
👉 1 to 100 podcast: https://open.spotify.com/show/6WAB5oOzco9oWrQmYmuFZ6

🧠 Want to stay ahead in AI commerce? Subscribe and follow along:
📰 Subscribe to the free Substack: retailgentic.ai
📺 Watch episodes on YouTube & Subscribe for updates: https://www.youtube.com/@Retailgentic

Creators and Guests

Host
Scot Wingo
CEO and Co-founder, ReFiBuy.ai

What is Retailgentic | Consumer Behavior & Retail Trend?

Retailgentic, presented by the ReFiBuy Media Network and hosted by Scot Wingo, is where Agentic Commerce meets retail innovation. We help retailers and brands prepare for the future of agent-driven shopping with actionable news, expert insights, dxis, forward-looking predictions, and practical guidance on Agentic Commerce Optimization (ACO).

Speaker 1:

We have proven this in our work with customers, that by understanding the behavior of the people who come to your site, yes, it helps if you know who they are, but you don't need to. You can actually generate the site experience that's relevant and effective for them to drive higher conversion, like you said. So behavior based personalization instead of identity based.

Speaker 2:

Welcome to the Retailgentic podcast, where we explore the intersection of retail, ecommerce, and AI agents which we call Retailgentic. Our mission is simple. We are shining a bright light on the new Agentic AI shopping megatrend which will disrupt the trillion dollar digital retail market. The stakes are high. Google, OpenAI, Microsoft, Perplexity, Amazon, Apple, Meta, and Anthropic are all fighting to be the consumer AI shopping agent winner.

Speaker 2:

Are gonna be some tough decisions for online brands and retailers to make. What happens to checkout? How should payments work? Could you continue to invest in retail media networks? What does the world look like when over half of your traffic is not from humans but from AI agents?

Speaker 2:

These existential questions are the tip of the iceberg. On this podcast, we're talking to industry experts, analysts, and influencers to get answers to these questions, identify trends, and best practices. This is Retailgentic, your podcast partner in navigating our AI shopping agent future. And now here's your host, Wingo.

Speaker 3:

Welcome to this episode of Retailgentic. My guest today is Rishabh, the man with one name. He is the Co Founder and CEO of FERMART. FERMART is an AI native personalization platform that sits on top of your e commerce platform and provides customized experiences for your visitors based on their behavior, not their identity. Previous generation systems, this is a big, big category in our world, this personalization platforms.

Speaker 3:

Previous personalization platforms relied on your identity. We spend a fair amount of time understanding that and why is it important? Because it turns out this relates to Adjunctic Commerce. There is a method to my madness sometimes. This is one of those rare times.

Speaker 3:

So, you will find out, all about this. And then on the, through that, conversation, I always like to pick up another founder's founder journey. So we pick up a little bit of that and some of the company building and where they are. We covered a lot in a pretty short amount of time, so I'm going to get out of your way and let you jump right in. Enjoy.

Speaker 3:

Welcome to this episode of Retailgentic. Today, I am joined by Rishabh. He is the co founder of FERMART. Rishabh, welcome to the show.

Speaker 1:

Thanks for having me, Scot. Yeah, excited to be here. We know each other through a lot of mutuals, so it's great to be able to chat retail and agenda commerce.

Speaker 3:

Yeah, yeah. Been at this whole game so long that yeah, usually if there's another person in e commerce I haven't met, we have at least two fifty mutuals. It's kind of crazy.

Speaker 1:

Then you start to think, how

Speaker 3:

have we not met? Like, how is that possible? But it happens. You you go to these trade shows, there's a lot of people around. Of We've probably been in an elevator together somewhere over the years.

Speaker 3:

But now we know each other.

Speaker 1:

Really true. Really true.

Speaker 3:

Yeah. But I don't know your backstory. Tell us your origin story. Walk us up to the founding of FERMAP.

Speaker 1:

So I have a somewhat nontraditional background. I grew up a physicist by training. So after college, I had the misfortune of spending a brief amount of time at a bank during the great financial crisis, so in 2008, as the banks were imploding.

Speaker 3:

Did you choose Bear Stearns?

Speaker 1:

Morgan Stanley, at the time was a big debt house. Yeah, it was big it was big enough that it survived. Well, it survived. Well, I survived, yes. But it And my experience of being in banking during the crisis was it's just not a pleasant place.

Speaker 1:

And so I went back to my longtime love, which is physics. And I ended up getting a PhD in physics, but was always like an entrepreneur at heart. So I tried to start two companies while doing my PhD. Both of them failed for various reasons. But the long story short of it is those experiences led me to taking a job in Silicon Valley at a company called LiveRamp.

Speaker 1:

LiveRamp is well known for being the largest provider of digital identity in the online ecosystem. So stitching together email to cookie to mobile ID and sort of helping people understand that, you know, different people are actually the same person having a journey across various touch points. And so in 2021, I finally left LiveRamp to start my own company with the idea that it's great that we're using identity based data and event data through LiveRamp, but why aren't we using behavior? So when... Because I, Rishabh, am not the same person in every context, you know?

Speaker 1:

When on Instagram and I'm like scrolling cat videos, I have more of a curiosity discovery behavior. And so when I see an ad, I'm in discovery mode. But when I'm on Google and I'm typing something, I'm in high intent browse mode, right? I'm like blue shirt from Banana Republic, and I have a very particular mission that I'm on, and so my behavior is different. Or if I'm even more intentful, like if I'm doing a search in ChatGPT, now I'm doing a bunch of research and I have extraordinarily high intent.

Speaker 1:

I have a lot of context. I have a lot of information about your product. And I kind of want you as the merchant to confirm what ChatGPT is saying. And then I just want purchase, right, upon making sure that it's actually what... Like, you know, this is no hallucination going on, basically.

Speaker 1:

But I have very... Like, the same person has different behavior. And that's sort of the story to what led to the founding of the company.

Speaker 3:

Yeah. At LiveRamp, what was your role? Were you in the e commerce vertical? Is that where you got a little bit of a flavor for what we do in our world? Were you horizontal?

Speaker 1:

Yeah. So the first three years I was horizontal. I ended up leading all of our customer facing technical teams. So sales engineering, technical account management, support, product expert, things like that. The second half, actually did completely unrelated things.

Speaker 1:

So I built their healthcare business, which is now a north of $100,000,000 business, their podcast identity business. So actually for ads served through podcasts like this one, How do you map the effectiveness? And then I built a privacy business. So actually pretty unrelated. Put your privacy

Speaker 2:

in your

Speaker 3:

teeth as a CEO slash general manager type thing. But that, yeah. Mhmm.

Speaker 1:

Yeah. Precisely. Yeah. So I had three GMs reporting to me. Yeah.

Speaker 1:

And sort of built three... Yeah. Three three businesses that ended up doing pretty well.

Speaker 3:

Was one of your schools PIN? Did you go to PIN? Or you just slid

Speaker 1:

the jacket there?

Speaker 3:

Okay. Yeah. Yeah. Was that a key school year undergrad? Okay.

Speaker 3:

Undergrad.

Speaker 1:

That was my undergrad, which is why I ended up on Wall Street for a brief period of time. Like everyone who goes to Wharton ends up at Wall Street for a brief period of time. It's a requirement. Yes, it is. It is.

Speaker 1:

It's a funny thing that they do up there on the East Coast.

Speaker 3:

Tell us the founding story. So did you and three other folks from LiveRamp leave? Or how did you find your co founders and what led to starting the company?

Speaker 1:

Yeah, actually, literally. I was just guessing. Yeah, yeah, no, exactly. Great guess. So Apple made an announcement in the middle of twenty twenty off the heels of GDPR.

Speaker 1:

So at the time, privacy was a hot topic. And so Apple made an announcement that said, You're no longer gonna be able to track people from app to website. That ended up becoming the whole, what in the e com world, people called iOS 14.5, but it was a policy called ATT that Apple implemented with the 14.5 rollout. And yeah, I was in the mix when those deliberations were happening because I was at LiveRamp and was in a senior role there. So I kind of foresaw the impact it was going to have on the commerce space.

Speaker 1:

And that's what was the sort of founding story is you know, the sort of like internet was shifting under our feet at that time because of privacy and then very shortly, soon after, because of AI. And it's like, okay, there's this massive shift in how the internet is structured and how consumers expect to be interacted with. And like, you know, you wake up in 2026 and you're like, yeah, actually, guess what ChatGPT does not do. It does not ask for your interests and your hobbies and your friends like Facebook did. It just learns based on your behavior and the context of your conversations.

Speaker 1:

And that's significantly more powerful by way of personalization than identity based data. And so the evidence was sort of self... It appeared on its own, that this is the way it's

Speaker 3:

in 2014, you weren't thinking about Chatuchipi Tea. Were you thinking about machine learning algorithms and kind of like that era of AI? I kind of put that in the AI bucket, but

Speaker 1:

everyone has In 2021, you mean? Like when I first started this company?

Speaker 3:

Oh, I thought you said 2014. So you...

Speaker 1:

Oh, my PhD was 2015. Yeah, yeah, yeah, yeah. Okay. I got it. And then this company was 2021.

Speaker 1:

And so we were at the late stages of ML, early AI. And then AI happened the time we had a public product.

Speaker 3:

On your website, let's help kind of frame what you guys do. And I found you guys on your homepage, you have kind of a framing that I found helpful, least. It's detect, recommend, and generate. What does that mean? And talk a little bit about these buckets that you have there.

Speaker 1:

Yeah, exactly. So the simple thing that we observed was that the old way of doing things was I would put a pixel on the page and either I would collect event data or I would collect the session recording, and then I'd have a bunch of humans watch this, like a subset of the session recordings. They would, as humans, infer what the person was trying to do based on those session recordings, create ideas, which is, like, recommend what changes should be made, and then send it to a development team to build the next iteration of the site. And so what we said is, okay, that can be just done autonomously. So you have all of the data of how people are shopping.

Speaker 1:

Like an oversimplified way of thinking about it is I take the session recording, I send that to a bunch of agents who were trained on commerce and what a certain interaction type means, those agents then recommend what to do and send it to a set of agents who can actually generate those experiences. So then it'll, like, actually generate those site. And so you can just build the site that says, hey, based on these recommendations, here's what this site should look, and then you launch it. And so instead of using humans to do analysis and recommendation, the humans are just guiding the agent infrastructure on what kinds of things the businesses are trying to accomplish, and the agents can actually sit there and watch every session.

Speaker 3:

Is your target customer a retailer or a brand or both? Like anyone selling direct, I imagine. Right? Okay.

Speaker 1:

Yeah. Is anyone selling direct?

Speaker 3:

Yeah. So the e commerce layer is the platform layer, be it a Shopify or Salesforce, a DIY, whatever, is kind of underneath you. And then you're sitting on the site. And as consumers come in, you're so in my older school era, we would call that I would call it a conversion enhancer. And then within this bucket of conversion enhancers, have on-site search engines, recommendations engines, and then you have a little bit of a bigger platform called personalization.

Speaker 3:

You may not like that

Speaker 1:

framing, but exactly right.

Speaker 3:

Is that what you guys do?

Speaker 1:

Kind of personalization. Behavior based personalization. That's it. So instead of identity based, it's behavior based.

Speaker 3:

Got it. Okay. So within that world, whole idea is instead of it being like a cohort, the best we could do probably in the LiveRamp era is, okay, I'm Chewy and I know I've got dog and cat people and you're going to dog people get a dog view. There's probably like a third bucket, which is both. But just simplify, I'm a dog person or a cat person, show it to appropriate, right?

Speaker 3:

But you're basically saying, Okay, has a border collie and a cat. I'm going to show them very specific things. Pretend there is more specific stuff to them kind of thing. It basically gets into like a one to one type thing because you're getting down into that layer?

Speaker 1:

I guess if the old world was one to many, so like one experience for many people, I would describe this as many to one. Because the thing is, Scot is not just Scot all the time. So like, and that's sort of the core insight is it's fine. Like, you know, as a marketer, like the marketing side has always said like, Oh, it would be wonderful if we could do one to one. But I think that was actually a lack of imagination.

Speaker 1:

Because the problem is that the individual is just not a static thing. Like, how I behave is just so variant on any given moment or any given day that you want to meet me where my behavior is, not where I am boxed. Yeah.

Speaker 3:

So today I may be a hamster person, and usually I'm a dog and cat person. I'm stretching my metaphor.

Speaker 1:

Yeah, one exactly.

Speaker 3:

Of my superpowers is to stretch metaphors into crazy places.

Speaker 1:

Yeah, yeah. Or it's like, Hey, I'm like in a metaphor stretchy mood right now. And so then like, you're, you know, what you would love to see in that, when you're in that mood is like something that kind of meets you where you're at when you're Got in that it.

Speaker 3:

So I think one of the companies that's been around in that space the longest is called Bloomreach.

Speaker 1:

Yeah.

Speaker 3:

So that kind of are you the classic David versus Goliath? Those guys are pretty big, you know, I think. I think they're like $3,400,000,000 ARR. They've been around for a while. You know, always rumored to be on the IPO list in our world, but never quite there or decided not to do it.

Speaker 3:

Is that kind of who you're providing an alternative to? Is that kind of how you think about it? Or obviously, you've probably got the Venn diagram never overlaps 100%. But is that kind of like who you kind of do battle with out in the trenches?

Speaker 1:

I think if we sort of fast forward two years, I think that the success case would look like people no longer... I'll take a step back because Bloomreach does a lot at this point. So I don't, I mean, for those who may not be as familiar with Bloomreach, Bloomreach, as Scot was mentioning, started off with this idea of personalized site experience, but then over time has like, they have an email platform, an SMS platform. They've added every e comm capability that you could possibly think of. And if you were to look at that total set of capabilities, what you would see is that there's an underlying thing that kind of ties it all together, which is like, there's a CDP.

Speaker 1:

So when you know your customer and you know that it's Scot, how do you take the next best action for Scot? And that's what they propose. And that's sort of like how I believe... I don't know that they talk about it in this particular way, but that's like, if you talk to their customers, that's how they'll talk about it.

Speaker 3:

And

Speaker 1:

I think that what FERMART aspires to do is that you don't need to be a known customer to the brand. And so there's like two buckets for any given brand or retailer, like customers I already know, and then I'm gonna do one to one personalization based on event and like Rishabh bought a shirt, which I categorize as like what was state of the art ten years ago and why Bloomreach has become a big successful company, because it was state of the art ten years ago. And I think our provocation to the market is actually you wanna know how Rishabh is acting at any moment in time. And actually, you don't actually care that it's Rishabh because his behavior is going to lead to more relevance to his purchasing decisioning than the fact that his identity is known. And so now you can go from the known set of consumers to the total set of consumers that are coming to your website.

Speaker 1:

And we have proven this in our work with customers, that by understanding the behavior of the people who come to your site, yes, it helps if you know who they are, but you don't need to. You can actually generate the site experience that's relevant and effective for them to drive higher conversion, like you said. So behavior based personalization instead of identity based. As a consequence of doing that, yes, the end state would be we would subsume Bloomreach's personalization business, but they have a massive SMS and email and CDP business. So they're...

Speaker 1:

Let's call it three, like, let's say if we use your numbers, their 300 to 400,000,000 of revenue will become 200 to 300 if nothing changes, but it won't go to zero, right?

Speaker 3:

Yeah. Yeah, they've got multiple. They're kind of a suite at this point. How, so the thing that's counterintuitive to me is when you said it's really the behavior and I almost don't even care about the person because you would think we put so much weight into the past behavior of the person. You're basically saying the current behavior outweighs past behavior?

Speaker 1:

I'll start with the story, then I'll tell you the data underneath it, and then I'll go back to some anecdotes, some real stories of how this plays out. The overarching story is Scot and Rishabh are perhaps completely different in the context of our interaction with a given brand. So let's, like, use Banana Republic. So it's like Rishabh has bought 20 shirts with Banana Republic and Scot has bought two. Okay?

Speaker 1:

And so, like, Banana Republic is sitting there thinking like, wow, these are, like, totally different people, right, in the CDP world. And now let's frame a real world situation. Each Scot and Rishabh are Banana Republic shirt in Google. So we each have the same intent in that moment in time, and we're each looking for something similar. And we click on the item that we are looking for because we are trying to serve a similar purpose.

Speaker 1:

We're on like a shopping mission for something. Maybe we're going to an event and we're... We happen to be going to the same event even though we're totally different people. And so the outfits we're looking for are similar. Totally reasonable.

Speaker 1:

Right? And so we're going in. The provocation is actually in that moment, and Rishabh's behavior are far more similar than all of their past actions. Far more similar. And the reason that they're far more similar is because of the behavior and the intent in that moment.

Speaker 1:

Now here's the data. Let me give you some data that is actually So we're in a

Speaker 3:

little behavioral cohort at that particular moment. Correct. Correct. Now here's the same page?

Speaker 1:

Will see we will see the same page. We will see the same page in this context. Because in this context, Banana Republic does not know that it's me coming or you coming. We're only caring about the behavior. We're assuming that both of us are using incognito.

Speaker 1:

There's no way to re identify, all of that good stuff.

Speaker 3:

So

Speaker 1:

it turns out that if you do this on a behavior basis alone, relative to doing it on an identity basis, you can see a 20% lift in the conversion rate. So I'll give you a particular example. Let's just say the way that I want to shop for a... We we we can we can use the shirts again. But the way that I wanna shop for shirts in that context is I want to have a heavy browse experience because I don't know exactly which shirt I wanna buy.

Speaker 1:

I just know the category of shirt that I wanna buy. And that's why I'm ending up doing a Google search the way that I am doing it. And I'm looking for a... In a particular category, something that fits my needs, and I want a heavy browse type experience. The choice to give me that heavy browse type experience, which is the way you solve that behavior problem, that that's where that roughly 20% number comes from.

Speaker 1:

That is roughly 20% more likely to convert than choosing to show me products based on my previous purchase behavior.

Speaker 3:

I got it.

Speaker 1:

My previous purchases are irrelevant to that context. Cool. And so that's like what ends up happening.

Speaker 3:

That's super helpful. Thanks for helping me understand that. So now I'm going make a logic jump, you can see if I get this right. So you have built this platform that understands intent and behavior at a consumer layer. And then now we're building an agentic commerce layer, right?

Speaker 3:

And agentic commerce, whether it's from Chat to BT or ultimately agents talking to agents, I think we're a while from that. But if you start looking at the protocols for all those that exist, they always come with an intent package. Right? And they have an authorization package and identity package, but there's always an intent. Like, I am looking for a Banana Republic shirt that is black in this price range, and it has to be this size.

Speaker 3:

And I want it in this Like the intent is all packaged in there. One argument I bet you'll make is you were designed for the agentic world because you're already based on what's going to be coming your way. Whereas the old world is going to get this intent package and be like, I don't know what to do that. I'm looking for Bob Smith because I kind of like him built around a CDP and an agent. You're not a human.

Speaker 3:

You're an agent. I don't know what to do. I don't have a CDP for agents. Is that kind of where the agentic piece ties in?

Speaker 1:

Okay. I have a yes and for you. Okay. So yes and... The way I think...

Speaker 1:

And I'm very curious what you think. But the way I think about agents is like, I think to myself, when is the last time I interacted with an agent in the real world to make a purchase? When I bought a house, a real estate agent. That's like what we do. Mhmm.

Speaker 1:

And it kind of turns out you give that agent a bunch of specifications, and they come back to you with a home, and they're like, I got it. And you're like, no, you don't because I forgot to tell you that I also like this and also care about this. You know... And so the agent is only as good as the feedback that it gets. It's only as good as the context that you give it.

Speaker 1:

It's only as good as its ability to iterate quickly with you. Okay. So why is this relevant? Why is this a yes, and to what you're saying? So you're absolutely right.

Speaker 1:

We care about the intent expressed, how it's expressed, when it's expressed. But what's more salient in a context like that is, yes, you want to return the right things based on what the agent is asking for. Yes, also what you are going to end up learning is what happens in the next interaction when the agent comes back and says like, Hey, actually, sorry, I need to update what I'm looking for. It turns out I also care about the collar. I really don't want a Chinese collar.

Speaker 1:

I really want a regular collar. So can you please help me with that? And it's like, Oh, actually, I really wanted a quarter zip. And it's like those updates are again exhibiting behavior, which means like now you have to think about how to return information, how to return products, how to show it to that person. What are the non expressed?

Speaker 1:

See, the thing about behavior is it is the outcome of both expressed and non expressed intentions. And that's like the thing that we think is going to be persistent and is going to be aggravated potentially in an agent first world.

Speaker 3:

In your Banana Republic example, if someone lands on Banana Republic and they're incognito and everything, how do you, you can't, do you look back through the where they came from to understand intent? Or is it you're just watching it on the site and you know, okay, you can kind of see what they start to do and start building intent. Like how do you think about intent? Is it a graph? In Rishabh's physics mind, what's the shape of intent?

Speaker 1:

I mean, you got all of it. You do all of those things. See, because the thing is you can understand... Okay, we're going to go all the way into the weeds. Let's just go all the way into the weeds.

Speaker 1:

You can

Speaker 3:

That's our audience. Our audience loves weeds. That's the number one feedback is like, More weeds.

Speaker 1:

Okay, great. Let's use a meta ad because it's even more obtuse, right? Because in a meta ad, you're in app browser, you'll never know who the person is. The session lasts thirty minutes. We all know what the hell is...

Speaker 1:

We know that we don't know who it is when they come in from an Instagram ad. Right? That's a fair assessment? Okay, great. Yeah.

Speaker 1:

So you as a brand have no idea when the person is coming in from an IG ad because it's in an app browser. However, here's, like, the sets of things that you do know that today you're not taking advantage of and that a company like ours would. Right? When you look at the ad that the person is coming in from, you can literally go to that ad and you can get demo information. You can look at past performance of that ad.

Speaker 1:

You can look at frequency. You can you can see a lot of things. And if you do your work, you can use that to understand a lot about what probably happened before that person chose and why they chose to click on that ad. What happens when they click on that ad? What happens when they land upon having come in from that ad relative to the other ads that go to the same destination?

Speaker 1:

So you can not only back calculate on a given, you can back calculate on differences and you can observe the differences in the behavior on the site when they show up. And then you can infer when the next person shows up, what is the likely path? Then you can confirm or deny that path very quickly within the first few seconds of the session. So there's a lot you can do without... Like, you can...

Speaker 1:

And I think most... When you say it out loud, it's kind of obvious, right, that this is possible.

Speaker 3:

So... But if you've gone down, there's kind of like an askew view, right, where if you're into the CDP world, you don't care about that stuff because you want to understand. I want their Facebook ID. Don't... Slash Instagram ID because I need that, or I need their IP.

Speaker 3:

You're trying to find these identity hooks along the way.

Speaker 1:

Right.

Speaker 3:

You're not really worrying about the campaigns. You may want to like a UTM so you can at least like give credit to the campaign, but you're not using the campaign for the contextual clues that it's going to bring downstream.

Speaker 1:

Precisely. And that's where my provocation on when you asked about, hey, how does this compare to Bloomreach? I just think Bloomreach was built at a time when the state of the art was to try to do it based on the identity.

Speaker 3:

Absolutely. Without getting... Just right there. Totally. Yeah.

Speaker 1:

Totally. And without getting... I'm not trying to get political, but doing things based off of identity was just popular at that time, whether it was in retail or internet or politics or whatever it was. That's how we did things. It became really popular.

Speaker 1:

And now it kind of turns out that you want to meet people based on their behavior in that moment, and that's actually more effective, whether it's retail or... You know what I mean? Like, it's it's kind of true across the board.

Speaker 3:

So... Yeah. I'm sure they would say we get 20% lift. You know, they're... Every...

Speaker 3:

This is what frustrates a lot of Sure. Merchants is, like, everyone has a 20 lift. So then you're like, if I got 10 of them, I'd have a 200, but then it doesn't doesn't work out that way. Is there... A fun thought experiment is, like, what if you did both?

Speaker 3:

Like, is there... Are they mutually exclusive, or can you do both? Is there a world where the intent piece at 80% and then the customer identity at 20% makes it 25% conversion? It does.

Speaker 1:

Yeah, it does. Because this is... And this was my point around the email and SMS thing. Yeah. Because like a thing that Bloomreach does or whomever else, Klavior, Attentive, whatever, like they will personalize the message in the email or the SMS that will increase the click through rate.

Speaker 1:

And then if what the person then lands on is both contextual to who they are and to the intent in that moment based on the campaign, the messaging, whatever else we have learned about how to think about the behavior of that person, then yes, it actually does compound. Now, of course, it does not compound linearly, but it compounds.

Speaker 3:

I want to pop out of the weeds for a second on a little side quest. And when I talk to some people in our industry about these types of solutions, whenever someone doesn't have one and they're shockingly large, I'm like, I thought kind of almost everyone had one of these personalization platforms. They'll say, my SEO team is worried about what it's going to do to Google bot, and it looks like cloaking or cloning. There's some SEO thing where Google has a lot of this stuff is urban myth. So I'm just I actually don't know.

Speaker 3:

I'm way over my skis on this. But there's something that they get really worried about. Like if Google looks down and they see one version of the site and then a consumer sees another one, that you'll get taken out of the index. It's kind of like the doomsday scenario that always comes up. How do you obviously, you wouldn't think that's true or you wouldn't be building this company.

Speaker 3:

But what's wrong in that talk track? And for the record, I don't believe that, but it comes up a lot oddly enough.

Speaker 1:

So I'll share two funny observations. The last time we got asked about cloaking was actually over a year ago. I think that ended up happening was... I think they just understood that there was a lot of things on their website that was dynamically changing anyway, and they just weren't sort of looking at it. And so the Google index saying, as long as it's aware of all of the content that exists, there's no problem.

Speaker 1:

And so this actually comes to the more technical point, which is we don't generate something that doesn't preexist somewhere or is cached somewhere that is retrievable by either our system in order to serve it or to the bot in order to look at it. And so it's not like it's creating something on the fly, because unfortunately, Gen AI is just not fast enough to do that yet. And so it is actually retrieving a preexisting page, content, whatever it may be. And so that's why it actually, from a technical standpoint, it does not end up mattering as much. So that's actually the technical reason.

Speaker 1:

Yeah.

Speaker 3:

Yeah, I think a lot of it goes back. I think this exists because there was bad actors, right? And the bad actors, their intent was to be a bad actor, right? So they were like, Oh, look, I'm selling, you know, candy for kids. And it was actually drugs, yeah.

Speaker 3:

And then you come in and it's, you know, or it's something else, or it's porn or whatever. And, you know, so there were black hat SEO people doing that. And then Google puts out this thing and then it gets, we have that where we like to put a lot of metadata on the page and the schema.org. And everyone's like, I'm going get accused of keyword stuffing. I'm like, no.

Speaker 3:

Keyword stuffing is when you say like snow skis, winter skis, cold rain, frozen skis. That's keyword stuffing. Or you like randomly put a bunch of locations on every page or something like that because you're trying to get location amplified.

Speaker 1:

Yeah. It's funny. I'd be very curious what you think, how you would react to this for what you all do. And I try to tell this to everybody. I'm like, Guys, your number one job is to say what's true in as complete, clear, and varied away as possible.

Speaker 1:

So that way, true information can be retrieved by agents and by bots and by humans. And so as long as you're following that, you're not going to get dinged because actually it turns out that those companies, so like Google, Meta, whatever else, OpenAI, they are using AI to verify what you are putting on your page. As long as it's verifiable and therefore true, you're not going to get dinged for it. You're going to get rewarded for it. Just don't say things that are not true.

Speaker 3:

Google is telling us, tell your customers to put this stuff in metadata because Googlebot will pick it up and we want it. And we want it in the data we want it everywhere. Like we want more data. We're like and then they're like, Google's going to turn and it's like, no, they're not. Okay.

Speaker 3:

All right. Let's zoom side quest over. Let's zoom back up to the weeds or back down to the weeds. The big thing in whenever I talk to a merchant now at that website layer, they're all adding a shopper agent. That's what I call it, an on-site shopper agent.

Speaker 3:

Is that an area you guys play in? Can you take your personalization? Or do you recommend they use a third party one and then you plug into that? How does that impact your world now?

Speaker 1:

Yeah, we can do either. It's funny. I was actually I don't know if I can can I share my screen on this thing? Yeah.

Speaker 3:

Yeah, you go into present or invite. Do have an audio audience as well as visual, so I will we will paint a picture with our words while we show the video watchers. We'll put some screenshots into the post.

Speaker 1:

Yeah, this is like an example. This is like an early release where like

Speaker 3:

Oh, this we're agent getting super secret stuff.

Speaker 1:

Is powered by... Well, I mean, it's on a public domain.

Speaker 3:

Oh, come on. It's an exclusive.

Speaker 1:

Come on, Rishabh.

Speaker 3:

Work with me here.

Speaker 1:

Sure, sure, But yeah, this thing is trained based on how, like on shopper behavior as opposed to just event behavior. And like, if you interact with it, I think you'll kind of like see, yeah, you'll see the differences in how that shows up.

Speaker 3:

You feel like having this based on behavior and intent is more, you know, I kind of know what you're going to say, but I'll tee you up. Do you feel like it provides better answers than kind of a more I could only know what the

Speaker 1:

other thing It depends on the off actually, my answer is it depends on the use case. And so I'll give you an example where it being just trained on facts is better. So support. It turns out that for support, you really just want to respond with facts based on the facts of the situation. So it's like, Hey, here are the facts of the situation, and you want to follow a set of policies, and you want to respond based on facts against those policies, against the facts that are given to you as a support...

Speaker 1:

That's the best case outcome, right? So I think the way that you train a support agent is very different than the way you train a shopper agent, right? Like a shopper agent, you actually like when it does some inference, right? Like you say some stuff, like you can try this. Like if you go to the website and say like, Hey, like I did this earlier today.

Speaker 1:

I'm looking for a bike to use in San Francisco. And it's like, okay, what you're actually saying is you're going to go up and down hills and that's going to be annoying. It's for commuting purposes. You don't want to be sweaty. And so it tells me what it thinks I'm actually saying based on what it can infer based on prior behavior of people who are doing the same thing and and then recommending based on that.

Speaker 1:

Right? And so...

Speaker 3:

It's like a little view into your inference engine from a consumer perspective.

Speaker 1:

Yeah. And that's like a very different thing than support. Right? And so I do think that it potentially leads to a very interesting question, which is I don't know how the future will unfold on this, but different use cases really do require different ways of training the agent. And you're gonna have to create some sort of architect, which is sort of going to your earlier question of like, Hey, do you care who people use or can they just use our data?

Speaker 1:

Ultimately, what I care is that they use our data. I actually don't care nor can I disclose who they use to power this shopper agent? But ultimately what I care about is that the behavior data is an important input into the places where you want to use behavior data. Yeah.

Speaker 3:

Is that agent part of your... Do you have a Lego block type package, or it's just like you get FERMART and you get everything?

Speaker 1:

The base thing that you consume when you work with us is the behavior data. And then off of that, there are use cases that live as like applications on that. And some of them you can use us for, you can use other people for, or you can just...

Speaker 3:

It's the Snowflake business model. I gotta, Everyone loves that one.

Speaker 1:

Yeah. And so we have like a sort of semi agnostic view of like how you use it. We care that it gets used, basically.

Speaker 3:

So give us a quick recap on the... We have a lot of entrepreneurial folks, I am a fellow, I'm a founder myself. Give us a little bit on company building. Think I saw on Crunchbase you guys have raised a pretty good amount of money. Where are you in that journey, and how big is the company, and where are going next?

Speaker 1:

Yeah. Gosh, I have so much to say on this topic. So let me try to think of the... I'll start with there's nothing more important than as the founder knowing why you want to build a business. That's the most important thing.

Speaker 1:

And if you don't have clarity on that, you're going to tie yourself up in all sorts of knots, whether it's a cap table knot or execution knot or hiring knot or whatever, you're going to tie yourself up in knots. So you just got to be very honest with yourself about why you're building this thing. Like, is it because you want to be rich? Just be honest with yourself and say that out loud. Is it because you want to be a founder and for some reason you think that's cool?

Speaker 1:

Okay. Then say that and be honest about it. Is it because you want to build a very large company? Okay. Why?

Speaker 1:

What's motivating that? And I think that the more honest you are about that, the rest of these decisions are easier to make. Because a question I'll often get asked is, should I raise money? Because people have seen we've done a pretty good job at raising money. And I'm like, I don't know how to answer that question without knowing why you're building this business.

Speaker 1:

Like, if you just want to become rich, then you for sure should not raise money to build a business. If that's the objective. Because there are way easier, way more higher likelihood ways to become rich than raise capital for a business. Now, if what you're trying to do is build a very large company, then yeah, sure, raising money is the right way to do that. But you have to have a point of view on either product development or go to market acceleration that will allow for that large vision being realised.

Speaker 1:

And so you really... It has to kind of be true that a large vision is realisable. And so I always fall back to like, why is the founder actually doing it to help unpack that. And then once you have a reasonable answer to that, then yes, there are tricks to every trade, right? But then it becomes easier to make those decisions on company building and, like, what are the risks that you should be willing to take?

Speaker 1:

What are the decisions you should be able to make quickly versus slowly? Things like that. So

Speaker 3:

That's that's very good advice. If folks want to learn more about what you're up to, obviously, in the show notes, we'll put a link to FERMART. But are you active on LinkedIn, or do you have a substack or anything? I think you mentioned you have a podcast. Tell us about that.

Speaker 1:

So the best ways to reach me are LinkedIn and Twitter, and I'm just Rishabh MJain on either of those. So that's just my first name, M for my middle initial, and then Jain, J A I N. Reach out to me on either one of those two channels. And then, yeah, we do a podcast on actually this topic, on company building. So we call it one to 100.

Speaker 1:

So it's like the journey from 1,000,000 to 100,000,000 in revenue, because our view is that that's the least discussed part of the journey. There's a lot of zero to one. There's a lot of very large company interviews. I think the one to 100 is like a pretty complicated topic. So we try to talk about that there.

Speaker 3:

I will check that out. Thanks for coming on Retailgentic, and really appreciate it. Good luck being a disruptor and building Vermont.

Speaker 1:

Thank you. Yes, thanks for having me.

Speaker 2:

Follow, rate, and share to stay up to date with Retailgentic, your guide to the future of AI shopping agents.