edited-audio-david-cost
===

[00:00:00] David, what's up man? Good to have you on the show today. Hey Jeff. Thanks for having me. I'm stoked to see you again. It was fun Last time we talked. You guys are running the most ridiculous setup of an e-commerce shop. I have seen in a really positive way in years of doing this. Rainbow shops. It's a huge retail operation.

You guys are running the entire e-comm op with like you and two engineers. From what I gather first, before we go deeper, how do you get up into running something like this? 'cause like the skillset you have to have to do some of the things you've done astounds me. I think just 'cause of the way you've run this.

I mean, so look, kind of career wise, I spent the first third of my career doing consulting, kind of a mix of management and technology consulting. I spent the middle third. I was a co-founder of a price comparison site in the very early days of the internet, so kind of the late nineties before Google existed.

I did that for about 15 years, and then I [00:01:00] kind of spent this last third taking. In that price comparison space, we saw the birth of e-commerce online and kind of took all those lessons that I learned from watching all of those E-com startups go from nothing to fairly substantial and said, okay, let me go take that skill set and go try it now on the retail side.

And that's what I've been doing for kind of the last third. And so you wound up at Rainbow Shops. You guys are running a sizable e-commerce operation over there. Just if you're in e-commerce, at some level, you're competing with the big Amazon Walmarts of the world. They kind of went the route of just hire every engineer physically possible, do a ton.

You guys kinda went the other way. I think you said last time we talked you had two engineers for the entire thing, but I just don't understand how that's possible. So we operate about 850 brick and mortar stores. Mm-hmm. As well as the online presence. We sell kind of affordable, trendy, fast fashion for women in both junior and plus sizes.

We also have kids merchandise from infants all the way up through high school. So that's kind of what Rainbow is. And what [00:02:00] Rainbow does. I describe Rainbow as, it feels like a TJ Maxx kind of format with clothing that in the old days you would've seen in a forever 21 or maybe something like today that you would see in a sheia.

So that's kind of the setup. You're right. When we ask our customer post-purchase, where does she shop for clothing, shoes, and accessories, other than Rainbow, same three answers we've gotten for the last two or three years we've asked this question, are Amazon, Walmart, and Chian? Huge, huge companies that we compete against.

And yeah, you know, we actually do that with two full-time engineers. They're very, very good and very talented. Apparently. You know, we use our two full-time internal engineers and then we partner with a lot of technology vendors who in some ways they're almost extensions of our staff. Yeah, so we're a big enough retailer.

When somebody builds something new, they need a place to test it. We are often willing to let 'em test it at Rainbow, and so in exchange for giving them the ability to test with us. We [00:03:00] have influence over kind of what they develop and how they build that, right? And so that becomes a way for us to supplement the stack.

You know, we were on Demandware or Salesforce Commerce Cloud for I guess the first, I don't know, 10, 11 years. I was at Rainbow. We re-platformed to Shopify a couple years ago. And why that platform decision is so important to us is. We rely on the broader ecosystem to supplement what we build internally.

So when we were originally on Demandware, Demandware was the king of the hill. That was the platform that everybody built for first. And so everything new that came out would come out on Demandware before it came out on Magento or another platform. We really saw kinda the marketplace shift to Shopify, and so we made the replatforming decision, moved to Shopify.

We moved in 21. But look, I mean, you know, we think today that Shopify is far and away the premier e-commerce platform with their scale. If you're gonna develop a new piece of tech that's gonna work in the e-com world. You're [00:04:00] gonna build it for Shopify first. You guys are building your core architecture there on the thing that if you need to partner or when you need to partner with other providers to stay ahead and just move quickly, you're basing your whole stack on the thing that they're all gonna be building on first.

So you're gonna just natively have access to kind of the cutting edge stuff. That's correct, and generally what happens is they need a big retailer to be able to demo or to have them use that technology, right? Mm-hmm. To prove to other big retailers that it works. You know, Shopify has a lot of small retailers, they don't have as many large ones, so we've kind of built a niche where.

We're willing to let people come try as long as they help or let us help influence that tech roadmap and influence what they develop and build. What does this actually look like in practice? Can you give us an example of where this has worked really well for you guys? Yeah, so we work with this company called Layers.

Mm-hmm. And here's what layers built for us. So the order that you display products on a collection or category [00:05:00] page is very important. Somebody hits your website, they go to your dress page. You need to show them something interesting in the first two or three rows. They're gonna get bored and bounce, right?

We may have 400 dresses, but the order we put 'em on, the page matters. Shopify outta the box has very basic sort orders that you can use to put things on the page. Not very sophisticated at all. In the Demandware world, where the Salesforce world, we had very sophisticated tools that let us build very custom sort orders to put things on the page.

We didn't have that when we went onto Shopify, we missed it. So we worked with this company called Layers to essentially help us build that same functionality and now the same. And again, in our case, we use this formula that takes the last seven days worth of dollars sales for every item on the website and then ranks or sorts products based on that dollar selling.

So we're using the wisdom of the crowd. To constantly change the sort order on the page. Interesting. [00:06:00] And I have to assume if you were to build this in-house, it's probably not the most complex thing to do, but then you have to maintain it. You have to anytime, maybe Shopify changes with two people, there's no way we would've built that in-house.

This technology competes against what an Algolia or a search spring, right? It's more complex than it sounds like, to maybe someone who's not as technical. You know, these companies need real infrastructure to be able to play this game wasn't something we were gonna be able to do ourselves, but working with layers, they built us a hell of a tool that gives us the same capabilities we had on Salesforce.

At a Shopify price and no, but here you are now talking about it publicly, you know, um, this, this is just a naive question on my end, but like, how long would that have taken to build internally, or it's just out of the scope of what you guys are looking to do? With an organization of our size and scale, we got a big brick and mortar footprint.

Yeah. But that doesn't mean we've got a big technology development footprint. It honestly, it would not have been something we would've built internally. Yeah. Obviously the brick and mortar side is pretty gigantic. [00:07:00] You said a number when we talked last time, uh, estimated number, and it was, I mean, your digital size is not small here, man, but you're doing it with two people, so like the economies of scale you're getting here are ridiculous.

Like engineer, engineer time is not cheap. They're both super talented and super productive people in the world of ai. AI has made them even more efficient. So as they rely on things like Claude to code more and more things that they want to get done, it's increased their productivity. We're getting even more out of the two of them than we were before.

This is everyone's dream. Every like founder's dream is, can I make this large digital footprint and large digital company with is like the minuscule number of a hundred X engineers. Yeah. And everyone thinks it's just gonna be a collection of agents or something. I do think you need the talent behind that to build it.

But you guys were doing this before it was cool. Instead of AIing a bunch of code, you just kind of worked with vendors for it. But this whole AI coding thing has brought up the whole conversation. [00:08:00] Of, is it the death of of software? Because who's gonna buy B2B tools when you can just build it so easily?

I think that's bs, by the way. I don't think that's gonna be true. I do too. I think these things look, everything looks easy until you actually start to code it. But you guys were already a great example of this, like focus on the thing you do and let other people focus on the things that they're gonna be great at and together.

It's gonna be fantastic. At Rainbow, I've run or been responsible for running a bunch of different AI experiments. I kind of put those experiments in two buckets. There's the bucket where the goal was to essentially replace people with the ai, right? So build a robot that could do what the human could do with the goal of reducing headcount.

Mm-hmm. If that worked. The second bucket is where the AI gives us some augmented or some superhuman capability, right? But it still is dependent upon a person driving it. Every experiment where we've tried to replace people has failed, and every experiment where we've gone, the augmentation route has been an enormous success.

In my [00:09:00] lifetime, I've watched a couple huge technology waves. So I came outta college. At the time, the first personal computers were coming onto the market. Mm-hmm. So the personal computing wave was an enormous game changer. We then get the internet, and surely after the internet we get the whole mobile phone.

I mean, I think mobile phone have been the big technological equalizer where personal computing was expensive and was more concentrated in educated or elites. Mobile phones make computing ubiquitous. Mm-hmm. So everybody has now a computer in their pocket. I think AI is the next wave and may be the biggest wave that I've experienced in my lifetime.

Of those three or four, you know, again, I was around when spreadsheets first came out and we didn't lose accountants or financial analysts or have fewer of them. Okay. We have more of them today than we did then. It became a tool that in the right person's hands could enable them to do amazing things, but it is a tool at the end of the day.

In our AI [00:10:00] experiments, AI is a tool. It requires human ingenuity, creativity, whatever, to get the most out of the tool. And I'll tell you kind of a quick story on this. We're working with a generative AI image startup called Leica, LICA. Yep. I mean, it's amazing what the people like have been able to do.

We're using them for product photography. We're also using them for editorial or things that you would see on a homepage or in an email. So when they were showing us the editorial tool, they had their own internal people produce some images to show us as examples that they were really proud of. They turned it over to our creative director who started playing with it, and when they saw what she was able to get the tool to do, even they as the creator of the tool, were blown away.

So in a super creative human's hands. They were able to get so much more out of the thing that they built than they were able to do themselves. Again, to me, this is yet kind of another example in this new world of ai, a human needs to learn how to use it and needs to embrace it, and I think people that do that [00:11:00] earlier will have a much easier career path than people that don't.

This is not a piece of technology that's gonna eliminate jobs. It's really not. It's gonna change roles and people that adopted earlier will do better. And look, I mean, you know, as good as the AI is, it can't end to end finish something on its own. It helps you get something done. It could take a lot of the busy work out.

It doesn't build the connections on the front and the back end. There are many things it can't quite finish on its own and it still requires somebody to tell it what it is they wanna build. Right. So back to Leica, speaking of like how you guys are able to operate on such a, a slim group of folks, so it's making all the product shots for you.

Like you have to have a source original shot of the product. Like what does this look like? Like how are you guys implementing this kind of into workflow there and what's it speeding up for you? Because this sounds like it could go wrong really easily. Here's another cautionary tale from the frontier of ai.

So it's interesting in some ways, the technology got a little ahead of our [00:12:00] ability to restructure the workflow to incorporate it. Basically what happens today, we receive clothing into the warehouse. We pull a sample. That sample goes to the studio where it's images need to be prepared. Comes out to an e-comm operations group that writes a product title, description, functional attributes about the item, and makes sure how it's gonna appear on the website, right?

And then something goes live. But even within that photo studio. It matters. Does it get shot on a model or not? Is it lingerie or swimwear that gets shot differently than if it's regular clothing? Is it a shoe, a bag, an accessory, jewelry? Those things get shot differently. There's all these processes and that's just getting it shot.

That's not when it's done. How it gets retouched, you know, before those images get approved to go live. Well, now you start to interject AI in it and you kind of asked what the input was. We've done two different kinds of inputs. We've taken very basic, just throw the article clothing on any person to just see [00:13:00] how it hangs on a body.

We then give that to the AI and it puts it on a model. We're also now experimenting with just throwing it down on a tabletop. So we're literally taking it out of the bag that it comes in, putting it on a tabletop, taking a picture, you know, from on top down. And then feeding the ai those, you know, kind of a front and back of that tabletop image.

And it is then doing kind of the composing to get that actually on a human. That's a lot of time of like styling it, setting it up. The human styling game has not gone away, right? So we still have to tell the ai, you know, we want this top, this bottom, this jacket, this pair of shoes, this bag, this belt. We need to tell it which things we want it to use to make the image.

It's not making those decisions on its own. Right? Again, not taking away the human work there, it's just, it's the taste that you can do things faster, but you still have to have the right idea of what to make. Now, you know, we're not gonna be sending, you know, images to an offshore place to be retouched.

We won't have to retouch images anymore. [00:14:00] Yeah, the image generation stuff has gotten wild. The pace that this progresses is unlike anything I've seen before. In this case, when we came back with some of those editorial images, Leica asked if it was something we shot in the studio as the creator of the tool.

They couldn't tell whether we shot it for real or whether it was AI generated. That's the phase we're moving into now, where if people are really honest and we give them two images, they're gonna have a hard time trying to figure out which is which. And I think in some ways it's really impressive and other ways, maybe it's a slippery slope.

I think the last frontier here still is all the AI video stuff We saw pictures were six fingers of people, seven finger people, until all of a sudden it wasn't. So it'd be interesting how that applies to e-comm, right? 'cause you can start to have the model walking or showing it in real life use, not just the still shot.

I'll give you my future prediction here. Has e-comm matures? Everybody kinda knows all the conversion rate. Techniques, right? We've done them all. We're all [00:15:00] looking for what's the next thing or what's new. I think AB testing image product images are gonna be the next thing when we have to shoot them in the physical world.

It's too expensive to shoot it multiple times. We can take an AI generated image, now we can set up showing the product in 2, 3, 4 ways. Mm-hmm. Randomly testing them on the PDP to decide which one has a significantly higher conversion rate and that ability to optimize images across your product catalog.

That's something that we haven't been able to do and I think AI will finally be the technology that lets us try that to build on what you're talking about there. The nice thing about in person is you can. Try things on. I can try like, do these pants go with this shirt? Does this jacket work? You can't shoot every combination of clothes.

Like you said, you have thousands of skews on, on the store. At our peak in the year, we'll have 75,000 skews left. Yeah, so that's a lot of thousands. Like you can't do every combo there, but something like this, if you can [00:16:00] start to render things in combination, you can sell outfits. There's an e-com retailer called a dormy in the lingerie space.

They understood the power of product images early on, and they built proprietary technology to be able to take so every piece of lingerie they shot. On two different models in two different locations, and then they would ab test every image and they found that one image oftentimes significantly outperformed another.

They had expensive enough product, limited enough SKU set to be able to fund doing double images for everything, right? You can clearly see this technique will work. I think AI will enable it. So I'm kind of excited in the next couple years where that goes, because I think once that starts, it will roll through the business.

Just like every other good conversion rate optimization idea does. One of the things for something like Log Rocket that we have seen is you can start to see how people are moving on your page. 'cause we do, you know, we started a session replay, but it's [00:17:00] also with the AI agent of watching your users. We can basically watch every session for you and tell you qualitatively about it.

You start to get into like this picture's different. How do people interact with the page based on that versus like, are they hovering, are they looking, are they popping through more carousels, right? There's a lot you can start to do with that kind of thing and feed that code or feed those suggestions all back in to some kind of tooling.

You have to ingest that and you can just start to rapidly really iterate on all that. For years, we've talked about personalization in the e-commerce space. Yeah. I would argue none of us have done personalization at best. We've done very crude segmentation. That's not personalization. AI with some of these agents where on your website you'll be able to essentially assign an agent, like a personal sales agent to each person who comes, and as they truly learn that person's preferences, they'll be able to advise, suggest whatever.

If this ab image testing thing works, you could even see that there would be differences from one person to another, and that those kinds of agents would be able to [00:18:00] pick up on that. Another technology vendor we work with is a company called Attentive. We do email and SMS with Attentive and Attentive is applied AI to do truly, in my definition, personalized messaging.

So rather than send a generic message that goes out to everybody, the AI is writing individual messages based on when the person last came to your website, what items did they look at, what did they add to cart? What did they abandon? It's using all of that data, all of those inputs. So on a welcome series where you know we're giving a coupon when you sign up or you've left something in your cart, rather than send some kind of generic card abandonment, being able to send truly personalized messages like a person wrote it is a game changer.

Oh yeah. When we first started working with them on these. And our ownership got one of the messages because they're constantly kind of playing with the website. They came back and asked who wrote the message because they thought it was a person who wrote the message. That's hilarious. [00:19:00] Right. So like there again, this is cutting edge stuff.

Using AI to do true personalization, we would never have been able to build that on our own. We needed someone like an attentive to do it. 10 needs a big retailer to be able to test it out with. Mm-hmm. So, you know, hopefully we keep building these scenarios where both the vendor and us win, which those are always the best partnerships.

So folks, you hear that if you can do, uh, super personalized photo generation, let 'em know. We could go on talking about AI forever. I'll be honest. It's the topic conversation so often when I talk to private people. Yeah. But another thing I want to kind of cover is, sorry to keep down the rabbit hole, but, but what you have done over there and what you and the team have accomplished with like so few people.

I keep using the word astounds me. Yeah. But I can't think of a better word. Can we talk about, you guys have a mobile app. You have a native mobile app? We do. Typically, you would need. A couple engineers just for that. And you don't again, 'cause you have two. So you guys use a company called Fuego and this really impressed me.

I thought this was a really cool story. Yeah. The people at Fuego have built something that [00:20:00] est what we've set up in Shopify and creates a native version for both iOS and Android. It has some capabilities that the website doesn't have, so it does take advantage of the things that you can do with an app that you can't do with a browser based thing.

Mm-hmm. But essentially the lift on our side is installing an app in Shopify and setting a couple customizations. That's it. So we pick up native apps along with push notifications, and in a world where we've already hit peak email, and I think we've probably hit peak, SMS push is kind of the next frontier.

No one gets in the middle of push. So on the SMS side, we gotta worry about at and t and Verizon and T-Mobile. Mm-hmm. You know, on the email side, we gotta worry about what Google or what a OL or Yahoo. So, you know, push notifications can be really powerful. Right. The person who downloads your app tends to be more committed to the brand than somebody who doesn't.

And so being able to offer native apps at a really low price and a really low, not only technical lift to [00:21:00] get it implemented the first time, but to maintain it going forward, I think it's a game changer. I'm always amazed at how few retailers have embraced apps. I've had apps at Rainbow most of the time I've been there, and I had it at the retailer that I was at prior.

My experience has been people in the world break into app people versus browser people. Mm-hmm. And you can't move people from one camp to another. Some people prefer to click that icon on a phone. It doesn't need to have any additional functionality, but they're just more comfortable with an app than they are, go into a web browser every time.

Mm-hmm. And vice versa. And again, in our case, that's about 20% of the people use the app. So that's a pretty big group. You know, if one in five people prefer to access rainbow product by an app than web, we really should personalize to them. They converted a higher rate, they repeat purchase at a higher rate.

Their average basket size is larger. Again, our ability to message them via push is invaluable. Lots and lots of benefits of having native apps, and again, by picking the right [00:22:00] e-comm platform in the right partner. And if wago. Lets us have something that, you're right. I could have had two people for iOS and two people for Android.

Right. We work with a lot of e-commerce vendors and the mobile app is one of the things that just, even the ones who have them, it's a giant lift and it's the thing that continues to like. Cause problems day in and day out because it's, you know, like you said, it's two different platforms just on face, right?

You have to have Android and you have to have iOS, but it's also just everyone's phone is a little bit different. Connectivity issues. There's just an infinite kind of list of variables that go into it definitely is. What I love here I think is at heart you have two engineers, but you also kinda have at some level, like 200, but you're only paying for two of them.

Yeah. I think in the world you have vendors and you have partners. Yeah. We've been lucky to overweight on the partner side and have fewer vendors, and those partnerships have been part of the secret sauce that makes rainbow what it is and lets us operate at the level of [00:23:00] efficiency that we operate at.

One of the big reasons we went to Shopify is because we thought their checkout only person that has to checkout better than Shopify is Amazon, and Amazon's checkout isn't for sale. I say, you can't use that one. So, you know, Shopify was a very close second. I think what they've been able to do with Shop Pay has been, you know, in the internet world we have problems that massive scale help you win at versus just having good technology.

A digital wallet like Shop Pay is, you need to have massive scale. Mm-hmm. And there is no other e-comm platform in the United States, at least. I don't know enough about the rest of the world, but there is nothing in the US that comes even close to the scale that Shopify has. That's why they're the only game in town for Shoppe.

Look at how hard digital wallet adoption is. So I can tell you on the Rainbow website, shop pay is up to almost half our transactions. Mm-hmm. So that's one and two people are using shop pay to check out. The next [00:24:00] one down from that is we might have an Afterpay or a Klarna at eight, 9%. A PayPal used to be 20 and now it's 10.

Even Apple Pay. Mm-hmm. Okay. With all of the power of Apple and owning the hardware. Apple pay is sub 10%, so what Shopify has been able to pull off at a 50 again, I think is miraculous and there is no challenger to that. Because nobody else has 3 million merchants that they work with, that they can automatically or push hard to run their payments through that.

If you understand how this works, anybody who's checked out with any merchant that offers shop pay, if you go to any other merchant that offers it, it's literally it recognizes you by email or phone number. It sends you a six digit code and you're done. Like this is as close to Amazon as you can get. You spend no engineering time on it?

That's correct. It reminds me of checkout here where like, just what are you doing? You can spend all the engineering time you want on it, and you will if you make your own. Uh, it's gonna be a constant problem. You're gonna have [00:25:00] engineers on it all the time. You need a bunch of 'em. It's gonna cause debt and it's not gonna be as good.

I think a lot of people in our industry, one of their negatives for Shopify was that you couldn't customize checkout enough. Look over my, you know. 15 or 20 years in e-commerce, checkout is not the place that you want to be an innovator. It's no different than in a brick and mortar store, right when you decide you're ready to purchase.

Our job is to get you checked out and bagged and outta that store as efficiently as we can to make you not wait or put up with anything. I don't care how fancy or how basic the point of sale is. You as a customer don't care. All you care about is how fast can I get in and out. It doesn't need to be fancy and doesn't need to look any different than anybody else's.

We're just taking your money and getting you out of the store. There are plenty of other places to put your resources in terms of innovating, being different. That's not one of 'em. At some level you guys are operating like, you know, the most innovative e-commerce company I've ever seen. And it's not [00:26:00] through Giant, well, it is through big pizza engineering, but it's because you've found smart ways to have partners, like you said, not vendors, but partners who are focused on the hardest problems in creating the best solution in the world for the hardest problems.

And you get to take advantage of all of it. Which is fantastic. Cool. Well, I could continue to talk to you forever. I'm gonna be honest. This is so freaking cool. I love what you guys have been able to accomplish over there, and I think almost any e-commerce company should look at this and just be just taking notes furiously from that conversation.

I'd love to catch up again in a while. Just hear about the new partners you found and the new ways you found to continue to innovate in this mall. 'cause it's so freaking cool man. I enjoyed the conversation very much so I'd be happy to do it anytime. Cool. We'll keep in touch and we'll have you on again.

But, uh, until, thanks for coming on, dude. This was a blast. I hope you have a good rest of your day. Oh, you're welcome. You too. Bye.