地下室から年商1,000万ドルへ：父娘ブランド Lilac Blonde のAI活用術

Joining me today is an AI commerce expert. At least I think he is. He sounds like he is. We had a chat before this. I hope I'm an expert enough. So why don't you introduce yourself and then we'll start. Yeah. Thank you very much for having me on, Mark. My name is Andrew Vitale. I am a co-owner of Lilac Blonde. So quick background. A little over 15 years ago, I entered the e-commerce world with a multibillion-dollar retailer here in the United States, replatforming their entire system. And we built that up to about $100 million using Oracle, Adobe, all the tools that you would think. And then I left them in 2017 into my own agency. But the more relevant part for right now is that my daughter and I started a women's.

Oh, yeah. Thank you. Working with your daughter is interesting and awesome at the same time. And unfortunately, she's not here right now. So we started Lilac Blonde in 2021. And what I really wanted to do is take my e-commerce expertise in digital marketing. It was an idea, a test, to see if women were interested in a sneaker retailer online. And I really wanted to get to know Shopify and how it could create campaigns and build that from scratch. We've grown over the last five years. We've done $10 million. This year, it clips about 60,000 orders. We're sitting on Shopify Plus. So it's been a lot of learning over the last two plus years. We've heavily incorporated AI tools and those pieces around marketing, analytics, and customer interactions.

So that's where we're sitting today. All right, let's break this down for the audience who are all marketers. So you're selling what exactly online with Shopify? So with Shopify, we sell our core customer, or ICP, is women 30 to 50 years old. They're based in the United States, so any of the 50 states. We are selling sneakers, which would be Nike, Adidas is our top brand, Doc Martin, Steve Madden. And then we added a couple years ago is really to complete the look, which kind of weaves into an AI tool. from the shoes up. So Abby's done a great job of bringing in clothing brands and really shopping and harnessing that desire for a shoe to fit a certain outfit. So our whole thing is how do we market to those women? And then not only for a new customer, but how do we grow that returning customer rate?

We are very deep with Meta, Google, Klaviyo, and Shopify. So we've done a lot of different tools in there. Okay, so is your value then, I look at everything above and beyond Amazon, right? So is your value is your matching shoes to outfits? Or, you know, why are they going to your site versus an Amazon? It's a good question. So I think there are two main reasons. One, we understand our customer best. We have developed our marketing, because that is my background, with a very crafted brand. The other thing is the products that we carry aren't your, as I say, regular degular shoes or clothing. So you may go to a finish line and get a pair of black and white Adidas ambas. we will carry not only that pair, but we curate and test in small batches, leopard print, an unusual colorway, double laces.

And then we take that data and we say, should we scale our exit? And then we continue to find a curated product. So think of it more like our goal was from the beginning, a boutique approach to mass market shopping for clothes and shoes. Okay. And Amazon doesn't provide most of that. Is that correct? No. Amazon is a completely different beast. Amazon is something that we've never really even approached. Our goal from the beginning was to own our customer data. And Amazon is you're paying for that customer, but it is very transactional in nature. So I really wanted to explore what it took to build the brand on Shopify and our own approach to building and owning that customer data.

It's important for us because we have a rewards program. We have Klaviyo, and then we have our own AI tool sets developed internally, and we've purchased multiple AI tools. So we can harness that data in a learning environment and really kind of shape what that customer experience should be. Okay, and those of you who don't know, Klaviyo is a CRM system, right? Or is it more of a marketing automation system? Klaviyo is actually email and SMS marketing, But like any major platform, you know, it's kind of like you move up the ladder if you're starting a brand and you really want to go inexpensive MailChimp, right? Yeah. Then we moved from MailChimp as the list grew and our segmentation grew to Clavio. And that is email, SMS marketing. But the thing is, every platform that we use has some sort of AI component now. so they have pushed additional services into that platform.

Okay, so let's get into your process. So let's say my daughter, since you work with yours, buys a – How old is your daughter? She's 16. Awesome. So she buys a pair of shoes. So tell me then what happens behind the scenes either with AI or your process for making sure that she returns as a buyer and that you understand her better than, say, an Amazon would? So I think that Amazon is a different thing when it comes to personalization and how they approach that product set. They're going to have a much larger catalog that they can craft a custom personalized experience around. Shopify does have those tools. So if you think about the Shopify ecosystem it has built not only with their platform but also an app platform So you can continue to enhance what you do from a customer service standpoint So I can add an app to turn on a banner

to create free shipping or to create a loyalty program. So the integrations are pretty deep and rich throughout that ecosystem. But for us specifically, we've developed a logistics system that started mostly in our basement. we tend to really kind of give a personal physical experience so we wrap the shoes in bubble wrap we write a personalized thank you card we have a little purple bag of emojis and little toys in there that was the beginning of our branding to create a one-to-one connection with the customer that feels like a brand and you're not buying from amazon uh shopify has the ability for us to not only manage the orders, but then, you know, to purchase shipping, track that. So that is all a baked in experience with the Shopify platform. But then from there, they become part of multiple customer streams.

So if they are put into our Clavio program, they become part of a welcome stream to kind of get them introduced more fully to the brand and they get email, SMS. then as they come back, the other thing is that we segment out our marketing programs, whether that is Google or it's Meta, that there is an audience segmentation for people that are not only browsers, but then people who are customers. And then you build lookalike audiences off that. Now, to varying degrees, both Meta and Google have shifted with AI this year, and we've seen ups and downs and how that works and doesn't work. Okay. Let me ask you a few questions about the site. Do you have AI deployed on the site to say they recognize somebody that's been there before, either through Pixels or I don't know if you're still using cookies anymore, or they've just signed in to your service and then you're making personalized AI recommendations to that user?

Are you doing anything like that on the site yet? Well, yeah, so there's two parts to it. There are the customer-facing aspects of AI, and then there is the back end. I'll start with the customer-facing. So we do pixel people. It does go through, I mean, there are cookies, but obviously since the change in iPhone adoption, that's not even a thing. But our first-party data does get shared to Google Analytics, Google Ads, Meta, so on and so forth. So we do manage that data from a customer experience standpoint. I think the biggest thing that you can see on the site, which we started two years ago with a company called Rep.AI, and this is an interactive chatbot that we're using for the parts of a shopping experience. It is a sales tool, so it can be interruptive in nature but trying to be helpful.

What are you looking for with this? What are the top things? It also is a customer service portal, so we can understand that data and then bring it back in and say, I'll tell you, for example, the number one question asked is how do they fit? So what we have continued to do is add sizing information, true to size, fits large, so on and so forth, and we surface that on the site itself. The other piece is our back-end application that we've been working on this year that will take all over, because it ends up being very siloed if you think about marketing, inventory, capital, and all that. And it always tends to be a lot of reports. So what we have done is really kind of tie those together with all of our first-party data and create an AI application that allows us to see, I'll give you a good example. We have the Adidas Samba Red Leopards on the site and Hot Cellar, right?

We've sold thousands of pairs of those. It will say you have 57 pairs. Well, when you have 1,500 items across the site, you do not know what are the highest volume sizes of those. Anybody who has sizes or variants of a product will realize this. and you can have broken size runs. So then you've got marketing going to a top-selling product because the AI is learning converts fast. But if you are not aware that that product is breaking, then you start to misspend dollars, and you see your cost per conversion go up, your ROI go down. So what we have done is our AI is constantly monitoring all the channels across the site, and then will surface to us immediately multiple things. You need to restock, protect. Or two, there's an opportunity here. I'm seeing good signals from customer journeys that this product is stocked healthy

and you need to promote it. So that's the back-end nature of AI that we use. Okay. Wow. All right. I'm sorry, did I give you? by the way for those of you that need more of a diagram like I do to walk through what he just said put on our YouTube page it'll be in the notes and you can see this animated if you're not watching from YouTube okay so very clear on what you do with AI on the interface itself what about any kind of AI I hate to call it CRM but But, I mean, customer relationship management. So you've got your best buyers or your regulars, your returners, your loyal customers. Is there any kind of AI being used to kind of analyze, you know, what makes them loyal and then try to make everybody else in your database loyal? Anything like that that you're doing yet?

Yeah, so that's actually a great example of how AI can be implemented. And I think that the biggest thing for people to realize is that AI is only as good as its source of truth. So if it is trying to decision off of third-party data that you haven't vetted, it could go a little haywire. You know, we were discussing this a little bit earlier today. What we do is we use our first-party data, customer information, everything. Make sure it matches. And then we have a loyalty program. So we understand what the lifetime value is. And we have KPIs for returning customer rates. So the way that we do that is we put our best customers in segments if they purchase more than once a year And it ends up being a marketing exercise But our loyalty program is something that we can continue to run in an automated basis.

Now, as far as an AI part of the loyalty program, they don't have that implemented yet. they basically, it is very much customer groups that you can send to. But everybody's trying to get up to speed from a platform basis and how AI and automation can make that material more actionable. Okay. And so AI is handling that and you're creating a solution that's not out yet, but that is doing all this just for Shopify users or for other things or other platforms? Yeah. So, you know, we fit in, as I call it, we're in the arena. We put to test all of our data. So we took, we have about 100,000 customer records, five years of data that we built in customer zero for this platform is really lilac blonde.

and we built it internally for us because Abby and I were having a hard time answering. Because if you think about how, especially a small business, but any business, I've seen it from big to small, it tends to be siloed that somebody's sitting at the end of a process and trying to analyze and know what's going on and then try to interpret that what it means for the business. So Abby and I actually started with business rules. How do you shape what AI needs to be with your data, with your business rules? Because that way, I think you and I were talking about earlier about how Claude can automate these videos. But you've created certain rules around that. You don't want to see things pulled in, a ninja turtle out of nowhere. Now it probably will. But it is your business rules as a base set. So the platform we've developed is within Shopify Plus sites, which there are 59,000 here in the United States that generate anywhere from $2 to $100 million plus a year.

And what we feel like is Shopify has Sidekick, in which we use it on the Plus platform, which the Plus platform has a very expensive cost to it for a small business. $2,500 a month may not be super expensive for a lot of bigger businesses, but it is for us. And what we find with Sidekick is it really is a way to automate reports and ask questions. What I have found over the last couple of years from a UI standpoint is it's kind of like if you're asked to take a multiple choice test, A, B, or C, versus an open-end essay question. And a lot of bots are sitting there, chat GPT, cloud, and all that, and they're asking, they're just blank. And you don't know what to ask or what results you're going to get out of it. And I think that's the toughest thing about most of these bots that are sitting there.

And I'm using bots probably, that's probably the wrong word, but agents, yeah. Because it can be intimidating. How do I know? And then people, I find when you do research within these, that people ask the same questions they were asking Google. They haven't really progressed, you know, how do I find a plumber? How do I buy this? So with us internally, we even sit there and ask questions. What we want to do is leverage in a semi-agentic way, how it could surface the information and be, for lack of a better way, and I know a lot of people use this, think Tony Stark, Iron Man, and the suit, Jarvis, responding to you. Or, for my age group, Kit, Knight Rider. So it's that kind of thing. How does it surface? What is important to you, and what is a capital decision that could improve your business right now? Yeah. Okay, and so what will it do?

It'll answer that question and go through the data? How does that work? Yeah, so the way it is is you, right now, you connect it to all your sources of data. So it's got connections for Shopify, Klaviyo, Google, Meta. And then it pulls in and synthesizes that data. And then it has a core set of business rules that you operate off of. So what we look at is a merchant that's selling multiple products, whether that is clothing or sporting goods or pet supplies. And then when it brings it in, it gives the merchant the ability to customize those rules to their business through a kind of WYSIWYG interface where you can say, no, we're fine selling something that's out of stock or whatever it is. And then it understands from those and can learn. So the biggest thing I found with AI is that I believe it's very commoditized.

And you and I were talking about costs, right? But there's so many tools out there that allow people to create, whether it's AI slop, we were talking about video and images, or just spin things out really, really quickly. So what you need to do as a business, or us, speaking for just me, is create a moat around your business. And I think the biggest things that businesses can leverage is the learning around their data, their customers, and what that moat builds. And then the application that we built internally was really to build, as people come into the site, we talked about the sizing pressure. If you've got a list of products on a landing page, because everybody comes to your site, has to land somewhere. And if you're looking for conversion or whatever, what are the products that are surface first to those customers? Now, you can do simplistic tools back in the day that's about personalization and all that.

What those myths is the sizing pressure. So if we are doing really well, what this application allows us to do is build collections on Shopify of products. then it will set them for you based on all the customer journeys it's seen, plus the inventory, plus your goals. So that way you can put that up on the page. So your marketing traffic converts better. So there's a lot of nuance to all that. I'm sorry. You're probably going too deep. No, we're not going too deep. But I think most people listening to this or watching this understand all of those nuances. and we like to kind of cut through them if we can and let people understand that there so many different use cases out there so many different industries and so many different reasons to use AI in your marketing So it nice to hear one perspective And somebody else will listen to this and they say well I could apply this to

what I do, but by making these tweaks. So we don't mind the nuances at all. But to wrap this up, we talked about image creation. And you're very particular about the images that are coming out. You want them to be as realistic as possible. I use ChatGPT. I've had good luck with that. You use Banana 2.0, right? Right. So I actually use for images, Nano Banana 2.0. I have, this has been a couple of years, I signed up for OpenArt.ai after exploring multiple different platforms. What OpenArt allows you to do is select different models for image and video creation. It's a credit-based system, so you have to pay. And I have that. Does it do it by sides? No, but you can create multiple variations of the same image when you put the prompt in or video.

Now, it's all dependent on how many credits. And I'll start off a silly project to test it. I want to create. We did a bracket challenge with all of our customers for the NCAA bracket. I love it. Yeah, it was fun. I lost, man. My dog almost won, though. Your dog almost won. Yeah, Abby didn't do so hot. That means it's random. Yeah, it definitely, yes, it is. But I think what you can find in there is that animated cartoonish-like graphics are very well defined, right? I can say, give me an in style of Calvin and Hobbes or whatever it is. What our challenge is, is I want imagery or video to be indistinguishable from any of our photo shoots or any of our videos that we do. and what we've developed is a SOP around how we build an outfit

and create imagery that showcases the product in the full look. But it needs to be, and this is where humans working with AI to fine-tune that. I was saying chat GPT can be a little cartoonish. Nano Banana 2 is the strongest. strongest. But you've also found a lot of success with Kling, right? Well, Kling 4 for video, as for realism, I mean, you can do celebrities in Kling 4. That's why I did it. I got a video I'm putting out that's kind of funny play on AI using celebrities. Whereas you can't do that with CDS 2.5. CDS 2.5 for me is a little bit better, more realistic. It's really I agree. And you know how I do it? I cheat a little bit. I just have Claude or Chachapiti write the prompt. I tell it the scene. It writes the prompt. And I get really detailed and nuanced with both of them. I test both of them to see which one comes out better. And that's the one I go with.

And it's damn good. With Kling 4.0, it's really good, too. And Kling is one-third the cost. So with animations, never use CD-Ans 2.5. You've got Kling 4. with real live video. If it's a celebrity, use Kling. If it's not, I'd probably stick with CDN's 2.5. I agree with you, and I think that I'll use ChatGPT as the opening. And my little tip here would be if you find, if you're creating images, let's say I think one of your previous podcasts talked about marketing images, if you find a style and things that work, save that prompt, OpenArt.ai works really well, and then edit the changes in that prompt. That way you don't have to keep on recreating the prompt every time, and it gives you more consistent looks. Well, I'm going to have a podcast on this where I've set up literally a Jarvis with GrokBots. I just have a podcast just on this where I've got five, six or seven bots that are talking to one another and all report back to Jarvis to me.

But what you can do is create a skill, and that skill can connect. I bet you it can connect to your open art, and then it'll constantly use that skill. So once you've found a look and feel that you like, you just use that skill that represents your brand, the way it likes things, and then it should do it each and every time. Yeah, the one thing I agree with you, and I think, you know, it depends on what level of detail for your business that you're looking for. I just did something for our client because I have a marketing agency they sell HVAC their service is HVAC plumbing so I had a little quiz created from a couple weeks ago that extremely well called Will It Flush and it's just a little interactive quiz that people take of stuff you can flush down the toilet it sounds silly but engagement's high but it created a great image but I had to kind of massage it.

But today I did Will It Grind for your dishwasher. And I wanted a Mario Kart-like theme, and Jatchi BG nailed it. So it is, and it didn't cost anything really for it to do that. The other thing that I would say is if you are creating AI images, is if it concerns you or not, Meta will tag a straight AI image as AI when you upload it to your, whether it's Instagram or Facebook. So for us, we don't want stuff to be tagged as AI. So there are ways that you have to bring that into something else because it's detecting the pixels in there. So it's just a little nuance thing. Yeah, I understand. Okay, well, I thought this was a very useful conversation. I know people listening to this will have learned something.

How do people reach you or your daughter? So, lilacblonde.com. They can reach us also at andrew at lilacblonde.com email, or at me at lilacblonde.com. So those are the main ways. Okay. And you're not blonde, so your daughter must be. She is, and purple has always been her favorite color. So we came up with that, babe. Yeah. I'm sorry. Just a dad joke. No, it was an awesome joke. And believe me, I'm deep into the girl-daughter world, and I always have been. So I can tell you a lot about American Girl Dolls. Fine. I bought a few myself. I know. All right, Andrew. So with that, look at the show notes if you want to get caught up or any of the contact information. And thanks again, Andrew, for being on. Well, thank you very much, Mark. I appreciate it. This is awesome. I love the conversation.