Episode 6

33 minutes 40 seconds Jun 4th, 2024

Reinventing Distributed Commerce with AI

Just as app tracking transparency rules shut the door on targeted ads, AI opened up new channels of customer engagement and personalization. Fermat is built on the convergence of those two paradigm shifts. The company provides AI-powered tooling and infrastructure for distributed commerce businesses to thrive in the post-ATT world. This episode originally appeared on Greymatter and is hosted by Greylock partner Saam Motamedi.

Rishabh Jain


Saam Motamedi:

Hi, I'm Saam Motamedi, a partner at Greylock.

My guest today is Rishabh Jain, who's the CEO and co-founder of Fermat. Fermat provides the tooling and infrastructure to enable distributed commerce. Rishabh, thanks for being here today.

Rishabh Jain:

Thanks for having me. Super excited about this conversation.

Saam Motamedi:

I've been looking forward to it, and it's been a long time coming.

Rishabh, I want to start, let's rewind the clock back to 2021 when you were beginning to think through (and then at the end of the year start) Fermat in the wake of Apple's App Tracking Transparency rules going into effect. The rise of ATT completely changed how companies think about ads and how they acquire customers. You come from a background before Fermat at LiveRamp really versed in all of this. I think Fermat was really born as part of that big transition.

Maybe to start, tell us a little bit about that and how you position Fermat in the new commerce landscape.

Rishabh Jain:

Yeah, totally. It's probably helpful to give just 30 seconds on what LiveRamp did or what LiveRamp does. LiveRamp provides all of the infrastructure to enable people to track across the open web. Anytime you're trying to find somebody and say that, "Hey, this is the same person who went to this website and then this app, and then did this transaction in store," it's like a medium-sized public company at this point. It was very successful in giving birth to this entire industry that now the entire open ad stack actually takes advantage of. That gave me a first class view into what it means to provide engaging consumer experiences. That's the way we thought about it at LiveRamp. What is the most engaging consumer experience that you can provide?

When you can't track people anymore, the entire premise of which we interact with the consumer internet completely changes. That was the core problem statement that I got very excited about.

Actually, when we first met, I was still at LiveRamp, but I was very excited about this problem and I couldn't stop thinking about this problem starting in the summer of 2020. Then we met at the end of 2020 and then the company started at the end of 21. Fermat's birth was to provide tooling, to provide that infrastructure layer that allows people to provide engaging consumer experiences on the internet in this new world of the internet. But now with other consumer internet changes like AI, I think it is even more obvious that consumer engagement on the consumer internet is going to be completely different than it was in the past.

Saam Motamedi:

Rishabh, I was just thinking back to when we met and you started explaining this all to me, and I came back to our team, like this is back in early 21 because at Greylock we've learned to start really large and important companies, there has to be some major shift in the landscape, and then you have to insert into that shift and write that shift, whether that's Workday writing the shift to cloud or Figma writing the shift to every company being designer first.

It felt like you were calling out that, as you just said, the way consumers interact with commerce globally and online was going to dramatically change. I think now three years later, sitting here in April 2024, that's proven to be true and then some. Maybe just riffing on that for a moment, how have things played out as you expected and also has anything surprised you in how it's played out around this shifting landscape?

Rishabh Jain:

It's a great question. I will share that I am thankful that you came with a beginner's mindset to this industry because I'll tell you that in early 21, when I was first doing discovery, I would call CMOs and ask them like, "Hey, what are you guys doing about the upcoming Apple changes?" The first set of answers I got was, "What are you talking about?" Then the second set of answers I got was, "Oh yeah, my agency and Facebook says they got it handled."

I could see in front of my eyes like, oh my God, nobody believes me. It was this very typical moment that a lot of founders have, unfortunately, that people don't believe you. Then of course at the beginning of 22, the changes actually impacted the industry. I'll be candid, I think that the impact was way higher than I expected, so much so that e-commerce businesses went from being money printing machines, which is why seven years ago they were venture funded.

I knew that it was going to be a high impact, but I did not expect that you would go from contribution margin positive to contribution margin negative. That changes everything when something like that happens.

And so you saw massive changes in the e-commerce landscape. But the high level is that you saw Facebook stock drop by 80%. You saw Shopify stock drop, e-commerce darlings got completely revalued. The way people thought about funding these businesses completely changed. The way people thought about operating these businesses has completely changed. So now, marketing and finance are closer together than they've ever been.

I would say that while the vector I had predicted was correct, I did not expect the impact to be as significant as it ended up being.

Saam Motamedi:

Yeah, I can tell you from the investor side of the table, we saw that impact play out firsthand in a number of commerce businesses that were pitching us and just seeing the unit economics almost overnight, just in the process of a couple quarters completely flip upside down. Then to your point, there's both the end businesses and then also the infrastructure layers and the acquisition layers and a Facebook and a Shopify.

What's the answer to that for brands and for people in the commerce ecosystem?

Rishabh Jain:

I think consumers are going to expect experiences that are continuous with the journey that they are on. The backend needs to support closed loop one-to-one experiences. The reason this must happen is when it is harder to target, you need the journey to be more effective. That is the simple thing. It turns out that when I knew Saam was the best next customer for me, I didn't need a perfect journey.

I could do with a bad journey because Saam was so predisposed to buying my product and I could just go fishing and find Saam and bring him to my website. That's no longer true. You need a great journey. Then the second thing is you need it to be closed loop. This closed loop idea is one that is completely underappreciated, and I'll share some of the ways this has played out in the ecosystem. The reason you need it to be closed loop now is when you can no longer track people, you need for the transaction event to be directly correlated to the intention event.

Facebook has done this by doing Facebook shops. TikTok has done this by doing TikTok shops. They have also done this by partnering Amazon and Facebook where you can now do an Amazon transaction inside of Facebook. Three years ago, that would've never happened. You would've never seen Facebook giving control of their feed to Amazon. That is a crazy change to the consumer internet that is totally under-appreciated.

It has also led to the rise of retail media networks, so you're seeing a very large set of companies, Instacart, Uber, all of these companies, you can see their revenue go up very meaningfully on the retail media side. Then there are companies like us who were fortunate to be creating this category where you can actually have a unique site experience for every piece of content that you deploy. Now the brand can actually own that consumer journey while still deploying it one-to-one to every piece of content.

We're very lucky that that has led to very, very fast growth and adoption because the need for these types of tools is so severe and you're just seeing the pull happen in every element, whether it's Facebook shops, retail media networks, or us. It has to happen.

Saam Motamedi:

Just as you were talking about the pool, I am reflecting on Fermat board meetings were always a joy because I come in and the message is always the same, which is "We thought we'd have a great quarter and we had an even better quarter than we thought," which I think part of that is credit to you and the team, but part of it is also I think a statement on what's happening in the underlying market and just the importance of this new tooling and infrastructure for these customers.

Maybe Rishabh, are there one or two specific stories you could share?

Rishabh Jain:

Yeah, I'm absolutely happy to. I would share two specific examples. The first is a customer in the health and wellness space called Mind Body Green. I remember there was a Monday at 10 AM, I just started to see new store creation notifications at a very high rate and something like 60 different stores were created in the course of an hour or so. I was like, what is going on here?

It was one of those moments where you saw the customer doing something that you knew at some point what happened, but it happened at a rate and in a way in which I could not have predicted. I learned later that what they were starting to do was think about customer acquisition not just as what ad and what creative drives a customer to do something, but actually it's a full funnel.

So what is the journey I want to put somebody on that leads to what business outcome for me? They started to use that data. They then pipe that data to the back end and they do a bunch of analytics to think about things like not only what is the best way to acquire, but also how do you think about what types of SKUs to push out? How do you sell those SKUs? What types of SKUs should go as single purchases versus repeat purchase and things like that. That type of thinking was exceptional. It showed me that you can actually not only retain strong acquisition, but you can actually start to drive your business in this new mode of thinking. That's one.

The other story that I can share is True Classic.True Classic is a very large men's shirt retailer basically online. We are starting to see ourselves get pulled into use cases that we truly did not expect. I'll hint at some of the types of things that we are starting to get pulled into. For example, some of our experiences with them don't even have a transaction. They have other engagement goals.

When you start to see yourself getting used in these new ways by these highly engaged customers who are at the cutting edge, that's when you can see that, hey, this thesis of the way that the consumer interacts with the internet or the consumer internet will change, and here's why. You can't think of everything yourself. Your customers will actually teach you how that ends up getting used.

Saam Motamedi:

Those are both really good examples. Earlier you said one of the big changes on ATT was one, it made it more important and harder on knowing who to target. Then two, how to make the customer journey be more effective. I want to focus on the second point, which is making the journey more effective.

You started the company late 21 in the birth of ATT. That trend's played out. Then 22 and 23, we had yet another trend, which is the rise of AI and the impact of AI across businesses. How has AI started to permeate into commerce?

Rishabh Jain:

Yeah, totally. I would say that the best thing about ads in e-commerce is that we are the fastest adopters of all new technology on the internet. For a different time, I think everybody ends up doing a tour of duty in ad tech precisely because the rate of innovation is so high. Speaking specifically about how it has impacted commerce, it has had impacted both the trad AI layer and at the generative AI layer. I want to talk about both for a second.

Facebook, for example, has been exceptional at what I would call traditional AI. The reason that they're exceptional at that is because it is actually a closed problem to find the best way to serve a given ad to the right person. It is a different type of problem to generate the first best guess of what is going to be an effective consumer journey. Those are two separate types of problems.

The way that AI will impact commerce is on both, but I think that people underestimate the total impact because of the investment in AI that traditional AI is also catching up to. I think that two things will happen in commerce journeys, and we're seeing it play out in practice.

The first is you can create more surprise and delight moments because you can generate a first experience that people could not previously have predicted. That's the benefit of generation. Then the benefit of traditional AI is that you can then continue to learn what are the things that make that journey effective for someone, and then morph the experience from there.

The way that this is playing out in practice is currently, you already have generative AI tools for ad creative. What you can see already happening is people are just running way more experiments on the number of creatives that they're launching, and then from there, iterating it down, generally using Facebook's pre-existing tools, into which are the ones that are most effective, and then doing the next set of generation after that.

We apply that same framework to the rest of the journey. What Facebook does for ads we do for the transaction journey. The beautiful thing is in our space so far, nobody else is actually thinking about that. There's all of these questions of what is the right way to set up the infrastructure so that way you can actually do that effectively and actually both generate surprise and delight moments for consumers, and at the same time deliver the best possible experience for them once you have created that surprise and delight moment.

Saam Motamedi:

Yeah, I love the way you frame that, and I do think it's important for all of us, given the enthusiasm and zeitgeist around generative AI, to also maintain the excitement around what's possible more broadly with AI, given all the investment that's happened beyond just generative use cases.

Let's talk a little bit more about everything that happens post the ad from a journey perspective. I'm a consumer, I see an ad for the men's T-shirt brand that you mentioned earlier, and I click through. What happens in a Fermat world?

Rishabh Jain:

It's a great question. Let's use the men's T-shirt brand as an example. This men's T-shirt brand also sells Polos, they sell crew neck tees, they sell V-neck tees, they sell hoodies. They sell different types of packages. They sell one of each, they sell a pack of six. Okay, so why is all of this relevant? Because as a business, what they're trying to do is they're trying to do one of two things. They're trying to acquire new customers, or the second thing they're trying to do is have the highest possible average order value for somebody who may already know about them, who wants to try a bunch of their products.

An example of something that happens when you interact with them through us is you see somebody wearing the traditional T-shirt. What they will then do when you click is they will continue the message of what attracted you to that T-shirt? Let's just say like, hey, it fits you super well. So they'll continue that content message and then they'll inventory products that are only related to that content message that attracted you there in the first place. Whereas the traditional website would inventory all of the products by definition.

Once you see the inventory of those products, you then select the one that you like. Let's just say it's a pack of three T-shirts. You can then change the messaging inside of the PDP to communicate about the fit over other benefits of this particular T-shirt brand because that is the thing that attracted you in the first place. The whole thing from impression to what you land on, to how the product description shows up to the cart…

Again, you can cart drive upsells using again, the types of products that are most likely for you to continue to interact with all the way through to the transaction. It is a concept that a lot of people don't yet... Again, our best customers know how to interact with this very, very well, but it's a new concept because it's rare that you would ever have a commerce journey where the product description page is different depending on how you got there. That doesn't exist.

The description is in today's world, but that's actually not how people interact with information. This is what we're seeing with Gen AI. This is again why I think Gen AI is important for a separate reason. It's going to retrain consumer expectations, but it is so important from a commerce perspective to answer the questions that are actually relevant to the consumer on that journey and to be able to drive experiences that do that because that is what consumers are going to expect.

Saam Motamedi:

Yeah, it's a great example because, and I'll just play it back to you, if I'm a True Classic, pre-AI, the concept that Rishabh and Saam both see and ad and are going to have radically different journeys based on who they are, what they want to purchase, it's an infinite set of potential permutations. I assume if I'm one of these brands actually building the development to support thousands and thousands of experiences, is functionally impossible because I'm essentially creating an unlimited number of potential destinations and journeys. AI is that superpower, if you will, that enables a company to work with someone like Fermat and automate that whole thing and create that really personalized out of one experience.

I think the beautiful thing, if I'm understanding it correctly, is you measure the efficacy of that in increased conversion, increased transaction size, and fundamentally stronger business. And so, it's also a use case of AI that directly connects to the KPIs that the business cares about.

Rishabh Jain:

Exactly. To expect anything different at this moment when people are talking about how generative AI is going to replace... The way that this is inevitable is that people prefer to ask Perplexity or ChatGPT or name your favorite AI tool a question because they can then ask the questions in the format that they want to, in order to get the information in the format that they wanted. That is for text. What is the obvious end state of that for commerce? It's what we're talking about. It's obvious and it is already happening. There is simply no way where you end up with a future where this is not what consumers expect and where the industry cannot supply that to consumers.

Saam Motamedi:

As you guys have layered in this AI capability and this AI-based personalization into the platform, what's been the effect on Fermat's business?

Rishabh Jain:

I would say the best thing about our business is we actually talk very little about AI to our customers. This is just something I believe in general about technology: any good technology is indiscernible from magic, but also the customer or consumer doesn't care. They need their problem solved. It turns out that people just want their problem solved and how they get their problem solved is irrelevant to them.

The problem that we solve for our customers is this idea that you can create a different experience for every journey that you are launching. We have solved that problem. We set up ambitious goals for this year, and I'm surprised by the rate of growth even given those ambitious goals that we have set.

We are seeing adoption that is very, very fast. We are seeing growth in actual usage on the platform that is very, very fast. If I'm being totally candid, it's like that level of growth when it happens that quickly, you then encounter a totally different set of problems, which is now the set of problems we have the fortune to be able to solve. You get this feeling that it's like, okay, wow, we are really able to help these businesses. We can meaningfully move metrics in a way in which people are very thankful and publicly thankful about how the impact is being felt, which then drives other customers that we get the benefit of being able to help, which drives the best employees in the industry to want to join you.

The net of it for me is I think that when you are creating a platform that will help change how people interact with the consumer internet, you want to do the maximum amount of net positive good for everybody who interacts with you. We are doing that for our customers. We are doing that for our employees, and of course, I hope we do that for our shareholders. That's how I am currently thinking about it given the opportunity that we have.

Saam Motamedi:

One of the things you said right now that really resonated with me is at the end of the day, the customer doesn't care about AI, generative AI. The customer cares about their business problem. If you can use generative AI to solve that business problem, fantastic, but you have to have clarity that what the customer cares about is the business problem. How have you navigated that?

I ask that because Rishabh, as you know, most founders building companies today, they're either AI-enabled or they're AI native. If they're not, they should really be asking why they're not. Then I think there's a question of, okay, how much do I lean into that in my customer positioning? How do I connect that to business value?

Rishabh Jain:

In general, I take the view that it is very important that when you say what your problem statement is that you have people nodding along, and people meaning your customers. If they're just like, "Yeah," that is such a pain. The problem statement I say is like, "Man, CAC on Facebook is totally ridiculous." And so it's like that's easy to nod along to. Then after that you have to convince them why your approach to solving that problem is actually a good approach, which is a separate thing.

You want people nodding along to the problem. I don't think that people nod along to, hey, what is the right way to use AI for X? I think what that does is that that raises questions. It does not lead to people nodding along. Positioning to me is more about first, how do you get people to nod along? Then it's about how do you get people to know you, like you, trust you, and then they will buy your solution.

When I think about company building now, I think it is a bad idea to not use the best possible tools to solve that person's problem. That's why I think it's very important. To choose to not be AI-enabled or AI native I think is not a great idea because somebody else could build your business and choose to take that route, and then they're using the best possible tools to build that business. That's actually the core issue from my standpoint.

Saam Motamedi:

Yeah, exactly. I think that connects to the question you were posing to me, which is the key thing is to have clarity on what's the customer problem and how you're going to solve it and whether or not you're truly solving it. Everything else can be iterated on. The reason why I say if you're not AI-enabled or AI native, you should really reconsider. It's not because that's the North Star. The North Star is solving the customer problem.

It's more an empirical observation that the underlying technology has gotten so strong, both from a generative perspective and a predictive perspective, that I think it's highly unlikely there aren't ways to 10X your customer value by leaning into it. Rishabh, to your point, if you don't, then you leave the door open for someone else to come.

Again, I think Fermat's an exceptional example of that just from my vantage point, which is again, if I rewind to when we partnered together in 21, for us it was like, hey, here's a founder who understands this domain really well and sees ahead and can see ahead of the next corner and is exceptional in that way. Then here's a really big shift happening to the interaction between consumers and the internet.

Candidly, AI was not part of the investment thesis, but I commend you and the team, which is like, okay, you're in these customer accounts, you're learning, there's a fast feedback loop. Then it's going back to what you said earlier, which is how do we keep making these consumer journeys more effective? Oh, well, there's a way we can drive radical personalization in a way the industry's never seen that I see it in the data is leading to performance and ROAS and characteristics on customers that are unheard of by leveraging AI.

We didn't start as an AI company per se. We started and still are a company focused on building the next generation of commerce. Turns out that we're leveraging AI to enable that, and that's led to a dramatic acceleration in the business and enhancement of the impact. That is just such a perfect example to me of what it means to have clarity around the customer problem, but an innovative DNA that enables you to co-op the latest and greatest from a technology perspective to solve that customer problem.

Rishabh Jain:

Yeah, I totally agree. One of the reasons why it's useful to ask yourself like, hey, what is the way that you're leveraging AI? If you have no choice but to ask how do you leverage AI that is going to be AI four years from now or six years from now, not the AI today? Because you're building a business for the long term.

The only way you can actually build compounding value is if in your system there is something that gives you a compounding advantage that future AI tools can continue to take advantage of. Then those businesses will actually accelerate even faster because the tools of the future are going to take advantage of those compounding benefits in an even bigger way.

The reason I like the litmus test of how do you use AI is because it forces the question of what is that underlying compounding advantage that other people are not going to be able to catch up to that will continue to allow tools like AI to drive leverage on your compounding advantage?

Saam Motamedi:

Yeah, I completely agree. Just to add and build on that, I think it is being in the customer workflow in a way where your product is tied to a metric or some form of ROI that AI can directly drive, and then also have the data loop where you're building a data set that compounds and enables you to get better and better.

In two years from now, three years from now, five years from now, what's going to be possible with AI in commerce and in the way consumers interact with the internet? How are you thinking about realizing that possibility through Fermat?

Rishabh Jain:

I saw on Twitter or X, a week or so ago..he's a very well-known technologist. He created this demo which was essentially interacting with a chatbot that felt like you were interacting with WireCutter, but it was personalized to you.

Basically it was like, hey, what is the best stroller recommendation for me? Then it would ask you questions to get to know you better because it knows the attributes of a stroller that are relevant to then provide the best possible recommendations. Then it gave you recommendations based on that. I'll tell you that the recommendations it gave me on the stroller question were very good. It's the same recommendations that I would give.

The reason I chose strollers is, as you know, I'm a parent and I did an incredible amount of research on strollers when we were having our kid. I consider myself like a stroller connoisseur. I was like, wow, this thing gave the same recommendations I would give. That's incredible. That level of feeling like I am interacting with a service or whatever it is that understands what I am here for and what my circumstances and therefore providing me with the best possible solution to the sets of things that I'm trying to solve for right now, is how every consumer experience is going to feel like.

You can take it from as a naive as a lot of consumers will have this, "Man, I hate this free shipping threshold. I have to buy something to make free shipping?" You can take it from as simple as that where you can actually start to manage against that type of objection or just fill that for the person just saying, "Hey, I know you're irritated by this free shipping threshold. Here's the way to get just over it."

It's like doing it for you for the things that you are expecting that person to want to actually seek in that experience, that's how commerce is going to feel, which today is totally unimaginable because we take so many things for granted. We take for granted that, hey, when I go into this thing, I'm going to have to pick it out of all of these SKUs. I think none of that is going to happen anymore. You're really going to have an experience that feels catered to the circumstance that you are in at that moment in time.

I think that the reason that that's necessary is not only because people will, again, get trained on that behavior because they're interacting with chatbots and Gen AI in general, but I think it's also necessary because that's the only way that you're going to build a sustainable econ business because the idea that you're going to be able to attract people is just getting less and less likely. It's like out of both necessity and out of consumer expectation, we can see the landscape quite clearly emerging in this way.

Saam Motamedi:

That's a great articulation and an exciting, exciting future for us to all look forward to. I'm excited to see the role that Fermat plays in pioneering that future for our customers and for commerce and consumer internet relationships more broadly.

Rishabh, thanks for joining us and sharing a little bit more about that.

Rishabh Jain:

Yeah, Saam, thank you for having me. This was really fun.