Originally aired on June 11, 2020 @ 4:00 PM - 5:00 PM EDT
Zack Bloom, Head of Developer Marketing, interviews Shalom Volchok (Co-founder and CEO of Outsmartly) ended up building one of the most tech-forward e-commerce platforms on earth, entirely with Workers!
So we are now live. I'm with Shalom, who is the founder of a company called Outsmartly. And this is an incredibly exciting conversation for me, because I think Outsmartly is the perfect, what I would call like an archetype of a startup. There's a lot of bad reasons to start a startup. You know, there are people out there, believe it or not, who start a startup because they just want to be cool or popular or get attention, or they look at other startups and they just kind of try to model it, you know, play dress up startup, LARP startup, basically. And that might even be most of the startups that exist in the world. But every once in a while, you have someone who actually genuinely legitimately has a problem. And that problem is meaningful, meaning it has impact on their lives or cost them a lot of money, something that is worth fixing, and also is uniquely solvable in the present moment. You could have had the startup idea of doing SpaceX, but if you happen to be born in 1900, that isn't a super relevant idea. But if you happen to have the problem at the exact moment when you scratch your head and you go, it is really ridiculous that, to use my analogy again, that rockets cost so much, because I know that the processor in my computer is more powerful than the processors in those rockets. I should be able to fix that. You have a convocation of a problem that is real, and technology that suddenly exists to solve it in a meaningful way. And you get a company that has legs that has a chance at really lasting and changing the way that people live and work. And for me, when I heard about UpSmartly, that is kind of the story that I see. Shalom has a lot of experience in the world of e-commerce that we're going to talk about, hopefully, and had a very significant problem that couldn't be solved with any of the platforms that existed. He ended up getting in touch with us because he needed to be able to solve it using edge computing, which hopefully we'll also talk about. And together, it forms a combination that I think has a shot at changing the way that people sell things online. Shalom, thank you so much for joining me. Yeah, absolutely. Thanks. Thanks for having me here, Zach. So yeah, so shall I start with the background a little bit? Oh, give me the story. Absolutely. Yeah, we'll start with the beginning. So my family had, of all things, a bulk herbs company that my parents founded in 1985. A bulk herbs company. So selling oregano. That would be more like spices. Selling like echinacea, golden seal, ginkgo, medicinal herbs. Early year of the Internet when selling stuff like that wasn't necessarily super popular and they had a lot of success realizing that there was a huge market for this stuff that wasn't being served. Yeah, this was way before the Internet, right? This was like 1985 is when this company founded, right? Oh, wow. Okay. Later on, in 2001, I joined the company and that's when we started getting into the Internet. But for the first, whatever that was, almost 20 years, we were like selling bulk wholesale medicinal herbs all over the world. Wow. So in 2001, I joined the company and we developed our first direct -to-consumer line of actually herbal cleansing detoxification kits. And I started playing with marketing those on Google AdWords, which is now Google Ads. And that ad spend grew to almost a million dollars a year over the next few years and it really revolutionized our business. So I built and managed and ran our e-commerce business for more than 10 years. And then after we sold that company, I wanted to get back into e-commerce and the intention was to run my own e-commerce business. And my co-founder and I, who had done some work with us at my family's company, we wanted to build an in-house platform focused around analytics and A-B testing and performance and personalization. You decided to sell something. You're like, we're going to go out there and we're going to sell scooters. And you realized that the actual underlying platform, you looked at Shopify and all the ones that I've heard of, and none of them were able to do the thing that you wanted it to do? Yeah. Really, it had come from when we had our company. The company was called Blessed Herbs. And there were things that I had wanted to do there that we didn't end up getting a chance to achieve. So it was really a continuation of what I had wanted to build at that time and wanted to continue to build. So actually, the technology came a little bit before the widgets and there were widgets. But in the end, the technology was more and more interesting, trying to figure out how to solve this problem. And for us, it really came down to how do we capture incremental revenue through optimization? Where optimization was three things, which is speed, how fast does it load for users, A -B testing, and personalization. And when we looked at the biggest companies like the Amazons and the Googles and the Netflixes, they've all been doing this since Amazon started personalizing in 1998. A long time ago. And performance has been critical there. I remember reading, there was a study that Google published where they had artificially delayed the response time for search results by like 200 and 400 milliseconds. I mean, teeny amounts. And they saw that the number of searches and the frequency of searches went down for those users. This was in 2009, Google was running these experiments, right? And interestingly, it got worse over six weeks and it took a number of weeks to return to prior levels once they stopped the experiment. So it has like a memory effect, like when someone remembers that your website is slow. So if I can reiterate those three categories, it's A-B testing, personalization, and speed. And it's this fundamental idea that anything you might be selling online, or in the case of Google, even just providing search results, if you're making $100 today, and you're able to reduce your speed, or do a better job of personalization, or tweak your content or layout or something with A -B testing, it's likely that you'll make $120 tomorrow. So there's some amount of room for optimization and improvement in all of these worlds. If Amazon is going to just show the same products to every single person, they're going to sell less widgets than if Amazon is able to show me the stuff that I'm likely to buy. And bringing that to every company is going to make them more money. So what was the problem? That sounds pretty straightforward to me. Absolutely. So the problem is that speed and fast page load times have been at odds with any kind of dynamic content. And unless you were one of these mega engineering organizations and could afford to go and deploy your own servers all over the world and figure out how to orchestrate running this globally distributed service, then performance really wasn't an option. And we look at what evolved on the open market in terms of A-B testing and personalization tools were standalone tools that mostly all started as client-side JavaScript SDKs that you install. And then they block the DOM from rendering while they pull in whatever the experimental variations are. And then once that's been changed, the DOM can render. So on mobile, this can slow page load performance by a number of seconds even. So this is really a performance anti-pattern, let's say. And then in the last number of years, we've seen more server-side A-B testing and personalization like API driven, which is certainly better. But now if you want to run those, you're still pretty much limited to a single origin server where that's happening. So you don't have the globally distributed application that is close to the end users. And that's the geographic distance. And also just that the Internet in getting from point A to point B is like the Wild West out there still. There's no telling how many routers and connectors and things, which is why CDNs, that's one of the things that they can do is help to route that traffic back to an origin. But it's still less than ideal. So where we would really like to have this dynamic content happening is at the CDN layer. So if I understand, just to reiterate the two bad options, which kind of if I'm an e-commerce company right now, it sounds like I, for the most part, have these two bad options. I can either A, use a client-side tool, meaning my whole website is delivered really, really quickly. But then before any users can see anything, it's all basically put on pause. And a bunch of code that I've written or someone's written runs in their web browser, does all the personalization, does all the A-B testing. And then some number of seconds later, my website shows up. So it works, but it adds an intrinsic and somewhat unavoidable delay, which as you talked about before, speed costs money in the world of e-commerce. Absolutely. Or I can try to do it on my server. And so I have a server somewhere that is serving my e-commerce site using some e-commerce platform. And in that server, I can say, I'm going to load some information about this user, and I'm going to decide to show them the products that are most relevant to them. But all of a sudden, what that means is when a user visits from Europe, that connection can't just be served by a nice CDN like Cloudflare sitting in Europe. It has to go all the way around the world, my origin in Virginia somewhere. And then I have to run a bunch of code that's going to do the personalization, and then it has to go all the way back. And that geographic cost of sending that request all the way around the world makes your website slow, and it gets slower the further away from wherever your origin, the person happens to be. And as the world becomes increasingly global, that becomes a higher and higher cost to pay. And so these are the kind of two bad options that virtually every e-commerce company I'm imagining has chosen between thus far. Is that right? Right. Yeah, absolutely. And with everything being SSL and secure, that adds additional overhead. There's more around trips. It's not just one trip from Europe to US East and back. It's a number of trips to establish the connection. And then on subsequent requests and all of this, it adds up. So where we would like to do that is we'd like to be able to cache, let's say, a default version of the page in the CDN, and then be able to personalize or alter that at the CDN layer. And then we're serving personalized HTML content to the user, which is going to give that great time to first paint. And then if it's a modern JavaScript application, the application can boot up and whatever. It can take over from there. But that first page view has the content that you want to be showing that user. And that's what we've been able to achieve building out on top of Cloudflare Workers. And if we go back, let's say, the progression of technology leading up to the existence of serverless compute at the CDN layer. If we go way back, it was like you had bare metal and you had your server and you were responsible for it. And then this progressed towards the cloud. And now you had virtual servers and you had multiple data centers. So in AWS, you could go and you could build a cluster and multiple data centers and you could use Route 53 to route your traffic between those. But the engineering effort to set that up, plus the cost to run all of those data centers, plus the ongoing engineering effort to maintain that, means that you have to have a huge amount of internal engineering expertise and a huge commitment in terms of budget to manage that. So that was like the generation just prior to this, if you wanted to solve the latency problem of geographic distribution. So when Cloudflare released Workers, we were very excited about it. And I think we were pretty early adopters in trying to do some of these things because we were trying to find out how can we run this stuff globally without having to run and manage, essentially, not the data centers, obviously that's given by the cloud, but still clusters and all of this. And that was the prior generation that we were even working on internally in our organization that sort of came before Outsmartly was like, how do we run our own clusters so that we can do full stack A-B testing where you're actually able to test different versions of the application, different versions of code, different versions of content. All of these things want to be tested, but it's very hard to build out the infrastructure to do that. To give people context, when we talk about A-B testing, what we're really saying is showing a different variant of the website to different people. And usually that's for statistical purposes, so that we can figure out which one performs better. So maybe we're going to have a buy it now button, or maybe we're going to have a learn more button. And we're going to see which one people click more and which leads to more conversions. But it sounds like what you're saying, Shalom, is you can go a lot further than that. You could actually have a new version of your site, or a redesign, or something that dramatically changes the nature of content of your site. And it's just as legitimate to want to A-B test that as well. Yeah, absolutely. And when I think about A-B testing, I feel like there's a lot of, I don't know the right words, I don't want to say confusion maybe, but maybe sort of misdirection on that A-B testing is somehow a tool to increase revenue. And A -B testing in itself doesn't actually do anything for improving the conversion rate of your site. Personalization does. That's an idea. Let's try personalization. Let's try this algorithm versus that algorithm. And now we're actually doing something that could increase the conversion rate. A -B testing is really a mechanism so that we can know if a change we make actually hurts performance or improves performance. And if we're just, let's say, going into Google Analytics and comparing today's conversion rate to yesterday's, or this week's to next week's, there's a huge amount of variance there. And I think that if we imagine we have a baseline and then we have maybe as much as even 20% plus or minus, when we start getting there, it's hard to see those changes just by observing day-to-day or week-to-week shifts. So everybody at every organization in the world pretty much, I think I can say this, is trying to improve the performance of their websites. Whatever it is. Everything we do is we want to make the user experience better. We want to increase brand loyalty. We want to increase conversion rates. That's what everybody's doing. And I think there's a bit of a disconnect between that action and then A-B testing as this thing that somehow magically improves it. It's not. It's really like if you look at the Googles of the world, it's a tool that they use because they realize if they don't have it, they don't have a way to know if the changes they're making are helping or hurting. And if you look at some of the studies from Google, they say like only 1 in 10 experiments, of course, they're very optimized, but only 1 in 10 experiments actually improve the metric that it's intended to improve. And a number of them decrease that metric. And at that scale, we're talking about less than a percent sort of shifts they're looking at can equate to millions of dollars of revenue for them. It's disheartening though because when an organization gets excited about doing A-B testing, they do one or two tests and then they're really hoping that they're going to get a positive and dramatic result. Because if you don't, you kind of go, oh, well, I guess we'll spend our time on something else. And so this idea that maybe only 1 in 10 of these tests was ever going to give you a positive and dramatic result, it means that people have to rethink their perspective on A-B testing, that it's not something you can just do a couple times and really fully experience the result of. Right. And I think what to me is, I don't know, it's sort of scary in there. Like I think back as a business owner, right, and we're trying to make changes because we want to improve performance. We want to improve conversion rates, right? And if somebody tries A-B testing, they do one or two tests, like you said, and didn't really do anything. And then they go back to what they were doing. It's not like they said, you know, changing the site does nothing. So we're just going to stop changing the site, right? We're just going to leave everything exactly the way it is. And the site's going to be this way for the next five years, right? I don't think there's a company in the world today that's saying that, right? If they have any kind of serious revenue, they're constantly trying to make changes. But A-B testing is really just a mechanism to allow people to see whether or not the changes they're making are improving those conversion rates or return visit frequency or these kinds of things. Right. And then when we get to the problem that Outsmartly is focused on, it's that page load times are critical, a known critical part of the user experience. Right. And there's many studies out there from like the Googles and the Amazons of the world that like it's 100 milliseconds matters. Right. And now we have like, let's take a typical Shopify store, right, where there you just keep installing apps and they're all injecting JavaScript on the page. And before too long, it's like there's this huge ball and chain that the site is pulling along and it can hardly even move. Right. And all of these things were supposed to increase conversion rates. But A, the store owners generally don't really know if they increase conversion rates. And then over time, if they have revenue, there becomes a fear. And this was the case at our business, right, of making a change. Like, we don't know, like, can we remove this thing? Because maybe it's going to... Yeah. You can't do anything because you're always petrified that anything you might do could basically destroy your business. Right. Exactly. And in the beginning, it's not like that. Because in the beginning, it was like we started advertising, sales started to come in like easy to see. And it was like, oh, where else can we advertise? Oh, over there. Oh, look, more sales are coming in. Right. And this was like that that went on for the first actually a couple of years probably of our business. But then it became harder and harder to know. And now we were spending a million dollars a year on Google advertising. We could have been spending more. We could have been spending less. But we didn't have any mechanism to make a case for that. And then making changes to the site, it was like I proved to myself that I had no idea what was going to make the site better. And once that realization sort of sunk in, it was like, OK, well, how do we get better? Right. And that's where it went to like diving into, well, how do the biggest companies do that? And that's really that, you know, A-B testing and data analysis and statistical analysis are frameworks to be able to, you know, make little wins over time that constantly move the needle up. But the thing is, without those kind of, let's say, bumpers in place, it could be that you make a change, it comes up, another one goes down. And maybe six months later, you've actually decreased revenue because you've made a bunch of changes that weren't wins. Maybe some of them were. And businesses, I think in a lot of cases, don't know. That aligns with my understanding, right? I think there are very few of these platforms, if any, that truly have all of the tools that you're describing built into them. Because when most e-commerce platforms were built, people just didn't take the stuff that seriously as they do now. It wasn't as much of a necessary component. I think one of the other interesting things you talked about is trying to build it for yourself and running into these performance problems. What I think not everyone realizes is Cloudflare is not just a bunch of servers deployed to like cloud providers or something. It's servers that are like inside Internet service providers in really remote regions around the world. You're not going to take your e-commerce platform and get a server deployed to an ISP in Mongolia. And it would never be cost effective for you to do that. Because if you have to pay for the whole server, and you're only using it for your e -commerce platform, and you only have X users in Mongolia, it's probably not where you should spend your money. But if Cloudflare does it, and it supports millions and millions and millions of websites, all of a sudden it does make sense to have that server there. And it makes sense to allow you to deploy your code on top of it, and to be able to run this e-commerce site from 200 different places around the world. So it kind of only works at scale, and it would never really make sense for any individual company who's not a Facebook or a Google to ever try to build a network with this level of breadth, I think. Is that fair? Yeah, I think that's very fair. I think it would be extremely resource prohibitive. Even with the cloud, I mean, you could build a smaller one, let's say an Amazon. But even that, even two different Amazon regions introduces a huge amount of complexity in managing an application there. And I think we're also seeing a number of other shifts that are really facilitating the market being able to have technology similar to the biggest engineering organizations in the world. And one of those is headless commerce. So headless commerce is essentially, you take a Shopify, and now in the monolithic version of Shopify, Shopify is actually rendering the HTML pages and assembling them and shipping them off to the user. In the headless version of Shopify, there's an API, in this case a GraphQL API, and you can now call that API and get your Shopify product data and interact and place your orders and all of that through APIs. So you've now decoupled the front end from the back end. And pretty much all of the existing commerce providers, whether it's Shopify, BigCommerce, Magento, they all are pushing headless commerce in the last couple of years. And then the front end is really getting eaten by React and Vue and Angular and these modern single-page JavaScript applications that allow the organizations that are adopting them to move faster, to build more unique, better customer experiences, really to be able to provide the best customer experience they can, which is pretty difficult inside of any monolithic, let's say, commerce platform, whether it's managed or open source, however it is. So part of what also enables what we're building to be possible is that we aren't running an entire e-commerce platform at the edge. That doesn't also make sense, right? So there's this decoupling that's happening where a lot of stuff doesn't need to run at the edge, right? So Shopify, someone can still be on there, right? And they can have Shopify managing their orders and their customers and their products and all of that. We don't need to migrate to Mongolia, right? We just need to migrate the pages and the HTML content that we want to serve those users. And then we need a way to personalize that content at the CDN layer. So that's actually a fairly, I don't want to say light, but let's say constrained problem. Like it's a solvable problem versus having to rebuild the entire stack where now it's like we have this monolith running on the edge. That would still be pretty difficult, but it's not needed. And I think it's actually architecturally a lot better to have this decoupling and to allow businesses to be able to choose like, hey, I want to use BigCommerce or I want to use Shopify and I want to use that with Contentful or with some other headless CMS. And I want to deploy my site here and I want to use React or I want to use Gatsby or I want to use Next. They can pick these pieces and build their application however they need to. And then where we come in, what we're building on top of Cloudflare and workers is this essentially layer on top of the CDN that is able to cache, let's say a default version from essentially any origin. So however they constructed that page is fine. And then we are able to provide personalization and A-B testing at that CDN layer. So how many e-commerce companies in the world right now are operating at that level of technology? Like are able to do personalization and customization in a way that doesn't require each request to go all the way back to their origin? It's only ones that are entirely building in-house. So these are probably getting into the billions of revenue. I mean ones that are really completely building in-house. Would be the only way to pull that off. I think so. I mean we don't see anything. Certainly, I haven't seen a platform that a company can just pick up and get that level of globally distributed performance that just works with their application. And of course, that's what we're working on building. How convenient. How convenient, right? Of course, right? But I think that it's inevitable that we're going to see more and more of that because we're seeing these increased capabilities coming from the CDNs. And I think Cloudflare is really the leader right now in terms of serverless edge compute. And it's exciting what's possible to build on top of that. So if I'm piecing it together, Outsmartly is a platform that I can buy and add on to any existing e-commerce site. And it essentially bolts the personalization and the A-B testing and the ability to deploy variants and all of these things onto a headless e-commerce site that I might already have? Yes. Yes. And even it could, Outsmartly could sit there at the CDN layer and help with migrations from legacy technology. So what we're building, could we even work with a monolithically rendered PHP site? Whatever, we could integrate with that. And then as they migrate, they could even migrate a portion of their application. Or they could be like, hey, we want to just rebuild the search application in our site with React. The rest of it's going to be served from this monolith. And now at the edge, we could actually inject that React search application into the site and serve that page. These are the kind of things that become possible. I kind of think of it as, you know what a service mesh is? Sure. For anyone who doesn't, it's the ability to have a layer that sits in front of all of your web servers and different kinds of servers and lets you control security and routing and all of these capabilities that you would have baked into them before. And now you can take out of each individual little service and you can put into almost a mesh that sits on every machine inside your network and imposes all of these things in front of whatever service you would otherwise be running. And it kind of feels like the same thing for your front end of your application where you now have a control plane where you can decide what shows up in a given region or how something loads or what variant of something is shown to what user on top of the existing application you might already have. How close is that? Yeah, I like that a lot. I think that's where it's going. And I think that it potentially solves so many problems that are difficult today. I mean, performance obviously is one huge one. But also if we go into data and analytics, there's more and more pushback to third-party analytics, right? And one of the things that could be possible is that we could allow pretty much any analytics to become first-party analytics because we have that ability now at the edge to route. To give people context on that, if I may, a lot of advertising systems work by storing cookies on users' computers from the advertiser's domain. So when you visit different websites around the Internet, they all load from the same advertiser and that tracking is able to follow you in all of those locations. There's a lot of pushback on that, though, because it potentially violates user privacy in hard -to-detect ways. And so there has been a reaction of making it so that only first -party cookies are allowed. It's only possible to track people on the actual domain that they're on. So I can tell if you visited Amazon yesterday, but I can't tell that you visited Facebook yesterday if I'm Amazon. And this is saying that it's possible to use existing ad systems, but in this new world where we're only tracking people on the sites that they're actually on because since we have this layer that sits in between what basically is a service mesh for the front end, it can route all of those requests out to those third-party advertising systems and then send them all back over the actual domain that the user is visiting. Is that right? Yeah, and that's probably a very interesting product in itself. I wasn't even thinking myself at the moment about third-party advertising cookies, but that could definitely also happen in that way. But even your analytics that you're using for yourself is typically like a segment or Google Analytics or whatever, is being served by a third-party domain, even if it's on your page and the cookie is being set client-side. Those cookies are short -lived today. The browsers are starting to delete them versus a server-set cookie that browsers, those still have the biggest priority. And the other thing that happens with loading that kind of stuff from third parties is that you have the hit of the SSL connection and the handshake and all that stuff has to happen again, whereas if you're able to call your original domain, it's all taken care of. You already have that. You are connected. So that alone could help to speed up the performance of a lot of assets just by allowing them to be served from your first -party domain so that the client doesn't have to introduce additional connections and additional handshakes to other domains. So I would love to, if we may take a step back and really understand how e-commerce actually works. So let's say you and I were going to start a business right now and we were going to sell desktop lights used to make video conferencing better. We decided that this is a really good moment for us to move a lot of desktop lights because everyone's at home obviously. How do you actually do that and what are these phases that a business goes through as it scales up to where you are at with a million dollars a year in ad spend? Yeah, sure. So I think about two stages, right? The first stage is what we call exploration. In the exploration stage, there's not really anything to optimize, right? You're comparing zero to something, right? So if you're only able to do 5% better than zero, you need to go home, right? That's not going to work, right? So in that initial stage, a company can totally see if something is working, right? It's like we need to go out and put ads here. Do the ads generate visitors? Do they come to our site? Do any of them buy? Do we have anything like a reasonable ROI? And then if you do, then it's like, okay, well, where can we go spend more money to get more ads, right? Myself, I like advertising a lot because I think that paid advertising is very inherently optimizable at scale because you have a lot of levers and controls over it versus like if a site is entirely dependent on natural search results, like certainly there are many successful ones out there, right? But that means you need to have one, a ton of content generation, and then you don't have so much control. It's still a little bit of a like black box in terms of Google could change its algorithm tomorrow and it does, right? It has happened and now ranking shift and it's more difficult to know what those do. And we can come back to search engine optimization because I think that's actually also pretty interesting and there's some compelling reasons. I mean, performance alone is now a ranking factor. So that also can have a big impact on the traffic coming from natural search. So to give context then, most people maybe don't realize how dependent on ads the average e -commerce business is. You know, we think of like Casper mattresses or something as like, oh, they got a nice, you know, mention in the New York Times. That's how people discover them or people Google search mattress, but there's so much competition in a lot of these areas that the real way most of their visitors come in is they pay $10 or $100 to get a visitor through some sort of ad and then that visitor hopefully converts to be worth $20 or $200 and the difference is the revenue generated by their business. And it's just much more scalable because if I wanted to grow a business with organic search, as you said, I'd have to produce a ton of content and each of those pieces of content are just going to generate kind of a linear, maybe hopefully constant amount of traffic. And so it's not really possible to exponentially grow my business in a reasonable way. Whereas if I start buying ads, I can just put more money into the giant machine and hopefully if the economics work out, more money will come out the other side. Yeah, I mean, I don't want to have someone who's listening to this and being like, our whole business is based on natural search. We're doing millions of dollars a year of revenue. Like what if this is crazy talk, right? I mean, those definitely exist, right? And natural search is hugely powerful, right? So who wouldn't want a million visitors a month from natural search? Everybody wants it, right? Of course we want that, right? I think the point that I wanted to make is like, because your question was like, we want to go out and we want to sell these desk lamps, right? And we want to get started. So I think if we're going to go try to write an article and hope that it gets ranked and hope that visitors come to our site, I'm not buying that, right? Like I think it's very unlikely that we're going to get anywhere, but we can go and we can write an ad and we can turn it on and three hours later, we can start to get data on whether or not like is our ad showing anywhere? Like if we're advertising on Google and we're picking search terms, like are people searching for desk lamps, right? And maybe they are, right? And like in our case, these herbal cleansing detoxification kits, there was a ton of people on Google searching for this stuff. And that was really just a, it was a stroke of luck for us that that happened to exist in like a right place at right time kind of moment. And then we just put those ads and it was incredible like year after year that just this stream of traffic came from it. If you go into like a Facebook and you want to put ads, now you don't have people searching for things that they want. Now you're trying to like pick demographics, like I want to target this group based on whatever their characteristics are and now I'm going to show ads to them and then maybe they respond or they don't, right? But the traffic comes to our site, converts or doesn't convert. And then if we're successful, now we're doing a lot of that. And let's say we have some millions or tens or hundreds of millions of dollars of revenue coming in. And now it's a whole different story, right? Now it's all about optimization because now we try something and it's like a little drop in the bucket. Like that's all we can make is little drops in the bucket and the drops in the bucket add up over time. But if we don't have a controlled way to understand them, we can't. You know, it's like snakes and ladders. It's like, you know, five steps forward and four steps backwards. You just, you don't have a way to know. And I think that one of these insidious things out there today is performance and page load times are really horrible. I think Google has said the average mobile load time is still 15 seconds, which is like ridiculously. Yeah, yeah. And I have seen like multi-billion dollar public companies with like 55 second time to interactive. And like, it just like, like on mobile, like these images that are like two megabytes, right? And you have this little screen and this image is like, you know, like we're on dial up from 97 or something. So it's pretty incredible how unoptimized the Internet is as a whole. And I think we're just gonna start entering like the early adoption phase, let's say of optimization, where now for the first time, the technology is starting to exist and a number of different pieces of technology, right? That enable pretty much any business to start to deliver a world-class web experience and certainly being able to deploy and run code at the CDN layer is an important piece of that. But it also has to tie in with like the modern JavaScript web applications and headless commerce and all of these sort of pieces come together. So let's say hypothetically that I am a reasonably successful e-commerce retailer. So maybe I built on top of Shopify. I started selling these lamps. I'm moving a reasonable amount of product. I have a, what you would, I guess, consider an unoptimized website. What should I, like, where do you start? How do you begin doing A-B testing? What are the things that you test or try? Do you just install out smartly? What do you do? Yeah, sure. It's a great question. So when we talk about optimization, we're talking about incremental revenue increase, right? So the bigger that company is, the larger that, like a larger 5% difference is going to be, a 10%, a 20%, right? If you're smaller, maybe it's going to be harder to justify. But certainly there's some line there where it's going to make, it's just going to make pure financial sense to go out and start investing in a modern technology stack, right? Like making the move to headless commerce, building your front end in React, maybe using Jamstack and static site generation, right? And being able to put together this very modern tech stack that is then going to allow you to like, go forward for the next five years, 10 years, and be able to iterate and build your business. And I think that we're in this transition period where the tools from the last 10 years that were provided to businesses to, let's say, increase the revenue, just start to have this horrible performance impact, which is every time you install a Shopify app, it's like, I've seen these Shopify apps, right? They load jQuery like five times and load Google Analytics three times, right? Because apps are pulling them in, right? And like, it's like, it's just, it's just, it's really a mess. So that worked in the last, let's say, 10 years because the Internet was new, right? I mean, like if you're the first one to go online and sell desk lamps and people want them, right? Like they don't care. Like people are just going to buy them. But I don't think that's, that's not the environment that we're in anymore, right? Today it's like people, there are lots of amazing websites out there. People are used to incredible performance. They're used to like technology that works. They're not like, like if we got thrown back to dial up, nobody would use the Internet anymore, right? It'd be like, what? Are you serious? Like you want me to do what, right? So I think there's like consumer expectations keep getting higher and higher. And I think it's at this point now where for a business, once they've crossed this revenue threshold and I don't want to like put a number on it, but I think there absolutely is this revenue threshold there where businesses have to make the investment in a modern technology stack or they're just going to fall, they're just going to fall even further behind. I think we're going to start to see like the marketers and the engineers and everybody who is working on this, that that's going to be like a must have on the resume is like how do we build modern technology stacks? How do we do these things in performant ways? How do we solve this problem of, you know, 45 third-party scripts that are just like death by a thousand cuts? I've seen this at many points in my career where the standards just get higher. Right. What was acceptable, hopefully this also happens in places like accessibility, what was acceptable is just not considered acceptable anymore in terms of best practices of what we do. A commenter just wrote in, a viewer, that the average Google Lighthouse score is 41. And that's out of 100. Right. Yeah. Yeah. Sure. Sure. It's going to be a while before all of these legacy and older systems are able to catch up. My fear, I guess, is that the technology gets so complicated that everyone just loses to Amazon because they are not able to keep up with the consumer's increasing demands and needs. Consumers don't want to put up with not having a personalized experience or having websites that are slow. And I see this in myself where, you know, my instinct is just to go to the platform that I know is going to work and I'm not going to have to struggle with. And so I have this kind of fear that we lose the race. And it all just ends up consolidated in a couple of companies that are able to actually deliver at this kind of scale. Do you think that that's going to happen? I don't think that it's going to happen. I don't think it's going to happen like that. No. I mean, Amazon is way ahead. I mean, they own a huge portion of that already. But I think there are always going to be niches and large niches. Like we were talking to an online tire company doing $300 million a year of online tire sales. I'd never even heard of them. They reached out on Twitter. They're trying to move to a Jamstack site. And I think there are an incredible number of businesses like that that, you know, they can compete in their supply chain or where they get them from, or maybe they're building something. I mean, what we had, we were building or making our product, right? It would only be on Amazon if we put it on Amazon. Right? But I think that the underlying problem that you're sort of starting to unpack there is that, I mean, it's a cliche to even say it, right? But the speed of innovation in technology is just exponential. It's crazy, right? So I think that what we see is that, you know, what's possible grows at this insane rate. And then what the average company is capable of can only grow at a much slower rate because it requires the diffusion of knowledge. And diffusion of knowledge is just inherently slow, right? Because it requires people to tell each other and to learn from each other. And like that process takes time versus technology just requires, you know, some very concentrated expertise to be innovating. And that like explodes. And I think that's another trend that we're going to see like set our company aside completely, right? I think this is just going to be a trend in the market is that we're going to see more and more companies figuring out how to close that gap or not really close the gap, but lift the companies so that now you don't have to have the expertise in order to utilize the cutting edge of technology. Whereas today, like largely companies still need to have like a really experienced engineer who has experience with modern web stack and all of these pieces in order to get there in order to get that hundred Lighthouse score. And like, we're working on a Shopify Plus store right now and we have a hundred Lighthouse score, right? Like it's totally possible today, right? Not to brag, right? But like, it's just to say that it's possible, right? That's really the point that I want to make, right? Is that like, how incredible is that that you can be using like Shopify and you can build your front end and you can get a hundred Lighthouse score. Like it's just possible. I mean, that means that anybody in the world could just go out and do that today. They could build that and they could get that except that we look at the diffusion of knowledge and it's still really, really hard, right? And I think that's where we're going to see a lot of innovation in companies in the next five, 10 years is figuring out how to make getting to the edge and staying at the edge, no pun intended, right? Just as easy as it is today to sign up and use Shopify, right? Like that's where we need to get to. And then businesses will be able to compete. Like the playing field, the technology playing field will be leveled in a way that it hasn't been since the Internet began. I mean, since the Internet began, it's been technology giants, they have completely dominated because they have all the engineering talent. They have the resources to deploy globally. They just, they have that. But now we get incredible open source projects like React coming out of Facebook. We're now like literally any company, any individual on the planet can build with the same front end technology that's powering Facebook. Like how cool is that? Or you can pick up Angular and do this, right? I mean, that's like, I think that's really a phenomenal development. And now you get like Cloudflare works where now again, like anybody can just pick those up and now they have this globally distributed application, right? And I think there's, you know, there could be a question of like, well, then where is there the space for a SaaS company to build? Anybody, if it's so easy to do, right? And I think that it's not that it's so easy to do, it's that it's possible where it wasn't possible even like one, two years ago. Like it's just what is now possible in the world is really the pieces for the first time ever are coming together, which is like serverless compute at the edge, headless commerce, these modern JavaScript frameworks, like all of these things that are really elite engineering output, right? Like it's as good as it gets that are now available. And now I think the next layer is how do we build platforms around those that enable broad access to businesses, being able to use them and deploy to them and even engineers first, right? Like it's still pretty hard to get a hundred lighthouse score. I mean, let's say forgetting the hard part, you need to put a lot of concerted effort on it, right? And if you're an engineer and you've got 40 priorities, maybe you just don't have like the space internally. Maybe you want to, probably you'd love to like go build a hundred lighthouse score, right? But maybe you just don't have the space to do that, right? And what if there was just a platform where like you just like, oh, I can sign up and now the platform does all these optimizations for me. And now I do get a hundred lighthouse score. Like that's awesome, right? That's not cheating. That's totally legitimate. I think we need more of that. I think we're going to see more and more of that in the coming years. So then I guess if I'm going to paint the opposite picture, if I'm going to paint the optimistic picture, the future is going to look like me saying, I want to buy a tire for my car. And instead of having to go to some giant platform named after a river in Africa and search through all these crazy search results that are all poorly organized and formatted and random pictures. And I can go to a site that's dedicated to selling tires and it's organized in a very intelligent way because they only sell tires. And I can find the exact tire that I need because they know all the tires for all the cars because all they do is sell tires. And then the experience that I have in terms of personalization and checkout flow and all these things that I care about as a consumer speed is just as good or is better than if I had gone to this crazy consolidated vendor. Absolutely. And I think that you touch on a very compelling sort of undertone there is that maybe the little guys, the little people have a chance to actually compete because they're able to focus, they're able to specialize. And I think that's what we're going to see. I think we're going to see specialization competing with the huge organizations that just inherently specialization is going to be their weakness by definition, right? Just because at that scale, how can you be a specialist in everything versus the purpose dedicated company, like you said, that knows everything about tires and that's what they do. I believe it. I mean, I shop at a long list of places that are dedicated to scuba diving or dedicated to audio equipment that sell things better than the giant companies do and organize things better. So I believe in that future. So I have one other thing I would like to talk to you about in our last few minutes, which is in the time that I've known you, and I don't know if this is still true, but in the time that I've known you, you've been running this company from Cypress. Yes. And living in the current world we do, where so much of us spend so much time at home, I think the idea of being able to start a company and run it from a beautiful Mediterranean venue seems pretty nice, to me at least. And so I would just love to hear what that experience has been like and what the pros and cons of kind of remote startup life is like. Well, sure. So I mean, in the last couple of months, we've been locked down like everyone else. So there hasn't been any beach or beauty, really. It's just been the inside of the office. But I think that we've been planning to have headquarters in San Francisco. And actually this year, I've been spending a ton of time in San Francisco and was about to relocate to San Francisco when the whole world shut down. Wow. So on our fundraising deck, it says like, the starters in San Francisco in scale team, right? And I was sort of thinking like, to go in there and be like, intentionally like just like cross that out and be like remote first. Because like, I don't know that, I don't know when the world returns to that place where maybe you have a small headquarters, but I don't know if we're going to see like a real large teams being co-located as much as we used to. And I mean, right now we have engineers in Los Angeles and Tennessee. Our CFO is in Salt Lake. I'm in Cyprus. I mean, people are loving it. I think it's now looking like an asset for us that we began distributed. And sure, now it's like, I'd like to be able to go out more, but in terms of work -wise, it didn't have an impact on us. So I'm a little hesitant to jump into like, hey, let's like scale a co-located team. I mean, I think we have to, it's going to be some time before that's like, I think becomes the norm for a startup again. What has it, I mean, if you were talking to it, you clearly made a decision to move to San Francisco at a certain point. So there was some sort of disadvantage to being remote and being in Cyprus. Has that gone away? Would you advise someone that if they were thinking about doing this, they shouldn't do it from a place like that? No, absolutely. I mean, do it from everywhere. I mean, I think just San Francisco is pretty unique as an environment in terms of so much is there that, you know, you can just bounce around in meetings in a day and it's incredible what you get to see. So, I mean, I like face-to-face meetings. Like it's a nice thing to have. So I think there's still a case to be made for establishing headquarters in a well -known place and having that status, let's say. But in terms of like under the hood, actual engineering and development effort, I don't know that there's... I think it's a much harder case to make today than three months ago, for sure. So it's sort of like we're looking at a future where maybe a bunch of executives and product people work in San Francisco and then every other human, you know, engineer, someone who actually wants a reasonable quality of life lives in Cyprus. Sure, or somewhere else. Right, right. We may see more of that. I mean, I think we will definitely see more of that. And I don't know if that will ultimately pull headquarters more and more distributed or not. I mean, I know like if you just talk about fundraising, that, I mean, it's... I'd rather go meet in person than be on Zoom any day. I mean, even like this interview, right? It would be great if I was sitting across the desk from you. I mean, it would just be... It would be more enjoyable. So there's always... I think that's the biggest drawback of remote is that three-dimensional human interaction is hard to replace. I have this persistent belief that we are very early in the history of the Internet and we are going to come up with technology. Like it has only become easier. You know, it's never become more difficult to collaborate on the Internet. It's never become more difficult to have this conversation. And so it's only going to get easier. And if we can just hold our breaths for a little while before you know it, it really will be possible to have some sort of simulacrum of in-person human interaction delivered over the Internet somehow. Like I guess no one has ever made money betting against the Internet. Sure, sure. That makes a lot of sense. Yeah, I mean, I think already today working over the Internet is there. I mean, I think all the tools exist and it's still nice to shake somebody's hand every once in a while. Well, we will have to see. It has been very nice getting a chance to talk to you today, Shalom. Thank you so much. Thanks for having me on. It was a lot of fun. Really enjoyed talking to you. Thank you.