Orchestrate all the Things - Andrew Ng offers AI for retailers with Netail. Featuring Netail CEO Mark Chrystal

Episode Date: November 29, 2022

Retail is big business. But like many other sectors it's undergoing a transformation, largely affected by the shift of consumer behavior from physical to digital. Many retailers are looking to analytics and AI to help them cope with the challenges. Andrew Ng, among the most prominent figures in AI, is now turning his sights to doing precisely that with his new venture Netail Founded in 2022 as part of Landing AI, Netail, a technology that enables retailers to auto-identify competitors across the internet and track their assortments, availability and optimize prices in real-time, today announced the closing of $5M in seed funding. We connected with retail veteran Mark Chrystal who is Netail's CEO to discuss the changing landscape in retail and Netail's offering.

Welcome to the Orchestrate All the Things podcast. I'm George Amadiotis and we'll be connecting the dots together. Retail is big business, but like many other sectors, it's undergoing a transformation, largely affected by the shift of consumer behavior from physical to digital. Many retailers are looking to analytics and AI to help them cope with the challenges. για να τους βοηθήσει να αντιμετωπίζουν τις προκλήσεις. Ο Andrew Ng, ένας από τους πιο δημοσιογράφους φίλους στην AI, τώρα ανοίγει τα πλευρά του να κάνει ακόμα αυτό με την νέα του έργο, το Netail. Υποθέτηκε το 2022 ως μέρος της ανοιχτής AI, το Netail, μια τεχνολογία που ευκαιρίζει τους εταιρευτές να αυτοδιακτικά αντιμετωπίζουν τους συμπεριφέροντες μέσα στο ίντερνετ και να διαδικαστούν τους ασφάλτες, την ευκαιρία και τις προσοχές,

και να οπτιμίζουν τις πράξεις σε πραγματικό χρόνο, σήμερα αναφέρε-time, today announced the closing of $5 million in seed funding. We connected with retail veteran Matt McChrystal, who is NetEl's CEO, to discuss the changing landscape in retail and NetEl's offering. I hope you will enjoy the podcast. If you like my work, you can follow Link Data Orchestration on Twitter, LinkedIn, and Facebook. I'm Mark Crystal. I am the co-founder and CEO of Nect of multinational brands like Victoria's Secret, Disney Store, American Eagle Outfitters, and David's Bridal. And in that career, I've led all or parts of every function of a retailer from production and sourcing and logistics and

warehousing to merchandising and managing the inventory and financials and being a head of e-commerce. I've kind of done it all. But always my passion was to apply data and analytics to drive decision making and competitive advantage. And so I've supported that passion with three master's degrees focused on analytics, including a master's in machine learning and AI, and a doctorate focused on consumer behavior in digital shopping environments. So really, you know, my background's always pointed at retail and application of analytics into retail. And recently, I took the decision to step out of inline retail because I saw a major shift happening in the space and felt that new solutions were needed. And so joined Andrew Ng at Landing AI about a year and a half ago to really develop cutting-edge AI software-based

solutions to bring into the retail space. Okay, great. Thank you for the introduction. And you've already kind of hinted that, well, the next question I had lined up for you. So I was wondering on the exact relationship between Nettale and Landing AI and well by extension Andrew Ng as well because it wasn't entirely clear to me. I mean I saw in the draft press release that I had the chance to view that there is some sort of common ground let's say but I wasn't exactly sure what the relationship there is.

Is NetEll actually a kind of spin-off from Landing AI? Yeah, that's exactly right. So Andrew asked me to come and join Landing AI because he was seeing and hearing, frankly, from CEOs in the consumer world, largely in retail, that they were struggling with a number of problems. And he wanted a retail and analytics expert to join his team and look at these problems. And so when I joined Landing AI about a year and a half ago, we started on this journey by looking broadly at the market and the problems these consumer-faced businesses were tackling. And we actually worked on solving specific problems actually with the largest retailer in the world. I won't name them, but

you can kind of guess. And also retailers who did like less than one percent of their revenue. And what we found was that all of these retailers across the spectrum were struggling with the same problem. And fundamentally, the problem is a shift of consumer behavior from physical to digital. And so we ideated solutions to the problem and tested and validated our ideas with various retailers. And ultimately, we found a solution that we feel is extremely compelling. It's actually developed or delivered strong results for our early adopters and we're now spinning off from landing ai to increase our ability to focus on the solution and then the complementary products that we'll we'll build around that core solution okay i see thank you uh all right so yeah i think few people would uh would argue against uh well the main premise of what you just said.

I mean, the fact that people are more and more going online for their purchases. I would just add to that a little bit that, well, maybe not everyone is necessarily buying online. And I have something in mind by saying that. So I'm sure that you are aware that well quarterly results were just announced and well the world's largest retailer I think it's probably Amazon just lost a significant amount of market capitalization and people have obviously been noticing and trying to to explain that and one of the interpretations that I saw flying around was the fact that, well, during the pandemic, mostly every purchase went online and that

wasn't going to be sustainable forever. And that explains the downtick. So I'm sure that people are gradually going online more for their purchases but possibly I think not everything will be online forever. Anyway the point I'm trying to get to I guess is that well I'm certain that this is happening you know we can argue whether this is going to happen to keep on growing indefinitely or not but it's definitely happening. So there's a number of other people that have noticed obviously as well. And so they're trying to tackle that exact same problem that you're trying to tackle. So what would you say differentiates what you do versus other solutions out there? Yeah. And actually, I think you make really interesting point about Amazon

and their recent results and what's happening in the space. And actually, to support your point, still globally, about 80% of retail sales happen in physical stores. So only 20% of the US, actually 15% globally happens through e-commerce. So it's still a relatively small percentage. And that's growing over time. It has been certainly that jumped higher during the pandemic. And we've seen a bit of a reset on that. What's more fundamentally happening, and it's been trending there for years, and we saw an acceleration of this trend through the pandemic, which hasn't really reset, is that consumers have really switched their primary decision-making journey to digital environments. So it used to be that if you wanted to buy a product,

you would go to your local stores, your local shopping center, you'd walk around. If you're buying a pair of jeans, you'd go and look at two or three different, you know, jeans retailers, you try them on, you compare them. And that's not how shoppers and consumers behave anymore. They go to the web and they do a search on, well, what's the right product? What's the right price? What are the reviews on the product? I think one of the phenomena that Amazon is struggling with is that actually this consumer is becoming more savvy about where to shop, how to shop online, all the options, and where they can do this kind of comparison of products. And so as much as anything, yes, there's a bit of a kind of bounce back to physical retail, but actually more fundamentally,

there's more competition online because consumers are figuring out where they can go and find these products online. It's not just Amazon anymore. Google Shopping is spending tremendous amounts of R&D to improve their platform. And so now we're seeing that the vast majority of decisions are actually made in these online environments, even though ultimately it then is driving customers to a physical location. They're not going to that physical location unless they know the product's available. They know it's the right product. They've compared the prices and the products against all the competition that's out there. And what's really kind of a major shift for retailers themselves, especially retailers that were born as physical retailers with stores and then grew into the e-commerce space, as opposed to Amazon, which is going the other direction. now make these decisions and whether they're competitive or not in the space. Retailers used to be able to go and physically walk around their stores and look at their

competition. Typically, a physical retailer would have five to 10 competitors that they would compete with head to head, depending on the location that they were in. And it was relatively easy for them to check, well, what are their products? What are their prices? What's the service that they offer? And then they think about, OK, now how do we adjust our model to make sure that we're competitive in that mix and that we have our own niche in that mix. The problem today for all retailers is that

the competition has actually grown over a thousand fold. And in large part, it's because the shopping is now done effectively through search engines and marketplaces and social media. And now everybody who sells a pair of jeans is now in that mix to win the consumer. It's not just the five or 10 stores that are physically kind of co-located together. And so to be relevant as a retailer, it is becoming harder and harder. And you can't compete the way retailers used to. Retailers used to compete on, well, where's my store located? Let's put it in the best place for foot traffic. Let's do great window signage. Yes, they'd send some emails and things like that. That really doesn't work anymore. And what retailers now need to figure out is, well, how does my price stack up against these kind of thousands of other options?

How does my product features stack up? How does my availability of products stack up? And how does the quality and not only the product, but of my services stack up in my reviews? Because all of that is laid bare for consumers now. And so it's become a real challenge for retails to figure out how to navigate that. And that's fundamentally where we've built a solution to help them where no solutions really exist today. Just to add to what you said, yes, I think you're right. That's what I was trying to get to as well. I mean, consumers may not necessarily purchase online, but they certainly do their research online. Right. Yeah, exactly. And that's where, again, retailers just don't have the tools to like price and availability and quality that I never

really had to compete in. And for consumers, it's great because all of the information on products that we want to buy is kind of laid bare. We can exactly see what works, what doesn't, find the exact right fit for us. But it's extremely daunting and challenging for a retailer trying to stand out now in this kind of ocean of products and data that's out there around the products. The flip side of that, I have to say, though, is that kind of small retailers and even retailers without actual, you know, like physical presence, except for a warehouse maybe where people can just pick up their deliveries from or from where their deliveries are sent to retailers, they can also compete if they're savvy online and they're competitive in pricing and people are happy

with them. I have seen that happen actually, and I'm sure every consumer has seen that as well. Yeah, and actually, I think that's great. It means that small business owners can really now compete and win customers, not only locally, but they can win them globally. And I think that's a really good thing. But if you're a traditional retailer, they're not happy with that because they're now having to compete in this, again, this much larger universe and they're struggling to adapt to that kind of new competitive environment. had a chance to see about the platform, the solution that Netail offers, it looks like pretty comprehensive package actually, like an end-to-end solution that has many different elements. So would you like to walk us through those different elements? Sure, yeah, happy to. So just to start off, so our core technology really helps retailers to, and frankly, retailers of all sizes. So we're happy to help the small businesses up to the large businesses, but really helps retailers auto-identify who their competition is across the internet.

So effectively, we have a solution that goes out across the web, finds for any particular retailer, finds who their competition is, and not only find who it is, but actually at a product level, we match the products to understand what's competitive head to head. And then we can track all of the competitors in terms of their assortments, their availability. We can optimize prices in real time. And we've seen really dramatic results in terms of improvements in web traffic and profits and revenue from the solution. And to accomplish this, we actually developed a breakthrough in web data collection that effectively makes the existing web scraping type of technologies obsolete. And so we have a much larger scale in that core technology and ability to capture data and track data in real time than anybody else in the market.

And then on top of that, what we're then building is additional services on top of that core intelligence, core competitive intelligence. And we're building effectively separate applications that leverage that data and extend into new capabilities. So we're looking at the price intelligence, assortment intelligence, location intelligence, which really attacks those key elements that retailers now compete on. And then also customer intelligence, specifically related to loyalty. And effectively, we're thinking about this in three pillars.

How do you attract customers in this new environment? How do you convert customers in this new environment? And then how do you retain customers in this new kind of digital competitive environment? And so we built solutions to attack that end-to-end, but all built on top of this core capability of being able to understand what's happening in the competitive landscape and using that data to inform the rest of the end-to-end solutions. Okay, thank you. Well, I have a number of follow-up questions on a number of different elements of this platform that you developed. Let's see, I would actually start by something not as eye-catching initially, but it did caught my eye and I'm going to explain to you why. So you talked about, well, data collection by doing web crawling, basically. And the reason that caught

my eye was because, well, I was under the impression that, you know, the score was pretty much settled on that one. I mean, to give you a little bit of personal flavor here, I remember doing that for a project assignment in my master's program. And that was a long, long time ago. And we did multi threads and we crawled the web and that was a pretty decent solution a long time ago and you know google has been doing that for for even longer i'm sure you know they have perfected the the art of web crawling so i wonder what possible innovation could happen in that space yeah so the the google certainly does it and their core businesses is built is built off that capability right. The problem is that for retailers, they don't have that capability. Yes, they can subscribe to services that to your point have been around for 10 plus years that do web scraping. The problem with web scraping is it's not scalable. It's

actually hard to maintain when the structure of the website is changed, it breaks, and you have to go and manually update. It also, there's really no intelligence in the web scraping in the sense of identifying, well, what's competitive with the products that I carry in my assortment? And so what we've done is we've actually kind of rethought that entire process end to end. And how do you actually go and get the data that, yes, you could potentially grab through web scraping in a very kind of inefficient way. How do you do it using AI in an extremely scalable way and also in a way that's extremely

intelligent for businesses themselves and get the data that's really relevant to them and their offerings, the prices that they want to bring to market, the products that they want to bring to market. And so in order to do that, you need to be able to go out and get a massive amount of data and then sort through that data to figure out what's relevant to any particular retailer. And we're using multimodal AI there, so a combination of computer vision and natural language processing combined, effectively looking at products the way a human would go and shop for products, looking at the images, the descriptions, the sizes, the weights, the different quantities in packs, all those kind of things are factored into whether this is actually a relevant item to retrieve for a retailer as a comparison point for them to make decisions against. And I think that's the big difference between traditional web scraping and effectively what we've built, which is a truly intelligent system. Okay, so perhaps then the innovation is not so much in how you collect that data, but more in what you do with that data.

Yeah, I mean, we've made a breakthrough in terms of the scalability of getting the data. Traditional web scraping, again, is not particularly scalable. So we've solved that problem. But then to your point, now once we have that data, how do we sort through and make sure that it's relevant to the retailer and absolutely a big breakthrough there as well? Okay, well, speaking about data, in order to be able to do

what you described, so sort of getting what's relevant for each retailer in terms of competition analysis, basically, I guess a fundamental requirement to be able to do that is to have the data about what the retailer has. So I guess you need pretty much their entire catalog in digitized form. So how does the retailer get that? I mean, do they get it from, I don't know, from the vendors that provide to them or do they have to digitize their catalogs themselves or what are the options there? Yes. So there are multiple options. So the retailers that we work with,

they can either upload their catalog into our systems and we can digitize it for them effectively. Or we can go and collect it. If it's available in a digital form via the web, we can go and collect it directly for them. Okay. When you talk about having something available in a digitized form in the web, where would that be coming from? So if I'm a retailer, I have a local shop here in Athens, for example, and I'm buying off a manufacturer in China. Would the manufacturer in China actually provide the digitized form that you talked about? Yeah, exactly.

So we've worked with retailers who are buying off factory lines in factories in China, for example. And what we found is that they all have a digital catalog in one way, shape or form. And so we can take that and ingest that into our platform. And then effectively what our platform then does is go out across the web and find all of the retailers that are selling exactly those same products, or our AI is smart enough to go and find products that are highly similar, that are likely to be competitive, that consumers are going to ultimately choose between and bring those back for the retailer as comparison points. Okay, I see.

One of the other points you touched upon earlier had to do with data sparsity, basically. So being able to develop algorithms using less data than is the requirement usually. And I'm guessing this is where the actual connection with landing AI technology and approach lies. This is something that Andrew Ng has labeled data-centric AI, which basically comes down to having data that's not necessarily big in volume, but of higher quality. And that allows the algorithm, well, that and a little bit of human intervention, I think, actually allows the algorithm to learn faster than usual.

That's exactly right. And so our technology is completely separate from landing AI, but the thought process related to Andrew and the data-centric AI movement is pretty consistent. And exactly to your point, we fundamentally believe that working on the data leads to better AI outcomes than just kind of a focus on the code itself, which has been the traditional mode of AI. And so in order to do that in our solution, exactly to your point, we actually bring the experts in to our process. So for any retailer

that we work with, for example, a retailer that might have 10 or 20 buyers or category experts for the categories that they manage, we actually have screens in our solution that allows them to go and see what the AI is doing, what the data is that we've collected. And we actually have been working with our users to design the screens in a very intuitive and easy way for them to actually give us feedback on whether the data is good that we're bringing back? Is the AI looking at the data correctly? Does the data that we're bringing back from the web correct? We take the inputs from those expert users and actually use it to clean up the data, to improve our AI models. And so the next time we go out and grab the next set of data, the next day the next hour, the model is actually learned from those category experts around what's relevant and what isn't relevant.

And over time, actually, the data and the AI becomes as good at scanning the market as a category expert would, but just at far greater scale and speed. Okay. Thank you. Well, what if retailers don't have category experts, or they don't have enough, or they can't spare their time to do that kind of work that you just described? We still have an AI that's getting trained. And so the AI solution that we have continues to learn from every user across the network. And so it's getting smarter every single day. So if there's a small business that doesn't really have the time to sit down and look at the data that's coming back and give us feedback, it's okay because our AI is actually out there learning every day anyway and getting better and better

and they will just benefit from those results. Okay. So then it sounds like you have a collective model, let's say, that learns from the sum of all the for larger retailers who want a more personalized model, we can go that direction as well and have capabilities on both. Okay, I see. There's some other parts of, well, intelligence for retailers that you apply that we haven't touched upon yet. So you also mentioned location intelligence. What does that do exactly? Yeah, so this comes back to the kind of fundamental way

that consumers are shopping now. And one of the key ways is, is an item available to me locally? And Google is a great example. They've spent a lot of money actually figuring out where consumers are when they're interacting with their platform, and then connecting that to the retail data to understand if the inventory is available within a certain distance of them, as an example. And it's a primary way then that the consumers are deciding, well, is this something I'm going to go and buy today or tomorrow

because it's available to me, or am I going to choose a different retailer because either they're the ones who have it in stock or they can ship it to me. And so the location intelligence that we are building effectively works with the retailers to figure out what inventory to stock in what locations based on the preferences of customers in those geographies. So it's not effective anymore for retailers to just stock every item in every store. They don't have the space to do that. And actually, assortments are largely expanding because of this need to compete online. So they've got to be selective about the edits they make. They're going to have a much broader assortment online and then an edited assortment that they're

going to put in stores and an edited amount of inventory that they're going to have a much broader assortment online and then an edited assortment that they're going to put in stores and an edited amount of inventory that they're going to have in each store location. And they need to make decisions on what inventory to put in those locations. And so this location intelligence actually gives them an understanding of the inventory that they should have in any geography, not just based on the foot traffic coming into the store, but also based on consumers searching digitally in that space and the products that they're looking for. Since you also mentioned the assortment intelligence here, I guess that in order for you to be able to offer that kind of service, you probably have to be able to integrate with and ingest data from systems such as inventory management and logistics management and the like.

So on the assortment side, where we're actually more focused is on the competitive elements of the assortment, not as much the logistics. It's more about what's the white space in the market and also how do retailers overlap with each other. So a good example is traditionally retailers watch market share very closely, but it's a trailing indicator effectively. They're watching it after the fact on a quarterly basis and they're getting that data from companies like NPD and Nielsen and others to understand how are they competing with similar retailers in their space. What we're building

is the capability to understand in real time exactly how the assortments of retailers overlap down to the individual product level and how demand in any retailer is impacted by the changes in products and prices in other retailers. And so the assortment intelligence is really focused on that. It's how do you stand out from your competition from an assortment perspective? And where do you have those heavy overlaps to understand, well, who am I really competing with head to head? You know, if there's a thousand people out there selling denim, am I really competing day to day with a thousand different retailers or are there actually, you know, 15 or 20 that actually affect my business? And those are the 15 or 20 that I really need to focus on. And so

our assortment intelligence figures that out in real time. Okay, I see. And I think the last part of those services that you offer has to do with customer intelligence. So what kind of analysis do you provide there? So in this space, we've been successful in deploying recommendation systems, but that's not something that's particularly new. We've just found that it's still a problem that retailers are struggling with. The other thing in customer intelligence, and we have a solution coming to market shortly here, is around loyalty analytics. It's a hot space in terms of retailers really looking to push loyalty,

especially with this kind of competitive environment. They're very worried about losing their customers to their competition and it's much broader competitive set. So how do you hold on to the customers you've already got? And there's a number of reports out there that say that actually, as we've gone through the pandemic, consumers are less loyal than they ever were. In fact, I think there's a report said 70, it was 75% less loyal than we were

pre pandemic, because, you know, the pandemic opened us up to actually let's just find the things that are available to us that limit our exposure to the COVID virus. And that opened us up to say, actually, there's lots of options out there and we don't have to kind of stick with the brands that we've always purchased from. And retailers are conscious of that and they're like, OK, well, how do we retain the customers? And there's a lot of loyalty platforms out there that effectively track points and, you know, do the tactical execution around loyalty. But there aren't a lot of really good solutions that actually analyze whether they're effective or not and how to actually talk to the customer on a more personalized level to get them to respond differently to different loyalty offers. And so we've developed a solution in that area that really does loyalty analytics specifically. Okay. All right. Thanks.

I think then by now we have a pretty good coverage of, well, what it is you do and the kind of services that people can get from your platform. So now let's shift gears a little bit and talk about, well, more about the business status, let's say, of where you are at the moment and actually the occasion is that I believe you are coming out of stealth pretty soon in a couple of days and also announcing a seed funding round so I was wondering what is the number of things in fact so I'm going to ask you a few questions and you can answer in any turn you like. So I was wondering what's the current status of the platform in terms of maturity and whether you already have early adopters paying clients? Yeah.

So we are, as you said, just coming out of stealth. Just closing a seed funding round led by Magarac Venture Partners in Pittsburgh and AI Fund, which is based in Silicon Valley, and then also backed by the Hong Kong and science and technology part. And what we've been doing for the last year or so is really working with beta customers, making sure that the solution we're developing is a good fit for the market. We have that validation and we are imminently releasing the kind of first production release where we can really start scaling out to retailers globally. We already have paying customers in Europe, North America and South America and we expect to now start expanding there as we come out of stealth mode but we also opened an office at the Hong Kong Science and Technology Park because we see this as a global problem. And we are going to use Hong Kong effectively as a beachhead into the APAC region. And so we've actually opened a small office in Hong Kong as well. Okay.

I wonder if you can share how big the company is at the moment in terms of headcount. So how many people do you have working on this platform? It's small right now because we've been in stealth mode. We haven't started the hiring ramp up as we come out of the funding round. And so we right now have about 13 people, mostly machine learning and AI engineers working on the solution full time. And now with the close of this round, we're going to expand that team, not only in terms of expanding the machine learning and AI capabilities, but also then hiring the marketing and the business operation teams that we need to scale and expand rapidly.

Yeah, makes sense. What about the model for the platform? I mean, the model under which people can actually use it because you do offer a number of services. So can people pick and choose and mix and match what it is that they need? Or do they get like one license for all? No. So effectively, we're building it as a core technology foundation,

which is this competitive intelligence across the web. And that'll be the core data platform on top of which we're building the price optimization, the assortment intelligence and the location intelligence. So we see those as actually four separate applications. Effectively, you can just subscribe to get the market intelligence without getting the price optimization, or you can now get the market intelligence plus price optimization, plus the assortment optimization, plus location, or any mix and match of those that you choose that you feel like you need. And then the customer analytics right now that I mentioned is actually standalone.

So you can subscribe to that separately. And it's all done through a subscription service. Yeah, I was kind of assuming it would be a subscription based because, well, that's really the norm these days. It's very, it's kind of, I think it's only legacy platform these days that have not already switched to that. Yeah, that's right. And I think retailers have now got their head around

that as the new model for software development and delivery into the space. And we think it's the most effective model for us and allows us to continue to iterate and build out the platforms that we want to build out. I hope you enjoyed the podcast. If you like my work, you can follow Link Data Orchestration

on Twitter, LinkedIn, and Facebook.