Yash Shah (00:00.078)
Hello and welcome back to Building Momentum, the show where we pay back the curtain on the exciting and often chaotic world of building a successful science business. I'm Yash, your host for the show where every episode we bring you the stories and strategies of founders who've been in the trenches, conquering journals, killing their teams, and building plans that people and businesses love. In this episode, we'll be chatting with Lin from vshop.ai.
is an AI platform that helps you 10x your model and product photos. We're excited to hear their story and the lessons that they've learned along the way. We'll be dissecting their wins, their losses, and everything in between. So buckle up, grab your headphones, and get ready to dive in the world of SaaS. Hey, Lynn, thank you for joining us for this episode. How are you doing today? Good. Thank you, Yash, and thank you for the invitation today. Absolutely. So let's start with this. Let's start with...
talking about the big problem that that we shop is trying to solve today. What is the challenge and who faces this challenge that you're trying to fix? Yeah, so essentially, we know there are like millions and not more ecommerce or transaction clothing retailers out there. And they all need images for their digital store or their offline stores. In the past,
It has been a struggle for smaller media merchants to be able to find the perfect model, the perfect studio or location for such shooting, especially for fashion and clothing. And we have been in the e-commerce industry for decades now on our team and we knew it was a struggle over the past decades or ever since e-commerce has been booming, but it was just something that we had to deal with as e-commerce merchants or small business owners.
And now with Genevieve to AI, this struggle, what is, you know, actual pinpoint was actually, you know, being able to solve using the new technology. So that's why, you know, since the new technology arrived and the 2022, that's when we started to think like, maybe this is something that we can actually implement and use with our know-how, obviously, and to solve this issues for
Yash Shah (02:14.35)
millions of small medium merchants out there and then and that actually got expanded into larger retailers or brands like Shein is using this because they have thousands of listings on the daily basis that they wouldn't be able to you know do everything manually or just hire models manually. So yeah so this I think it's it's very much a top ask.
for all of the merchants out there. Yeah, and I would agree to this. And the reason why is because I have a friend who runs a jewelry e-commerce platform. so those are sort of, know, in India, we are big on jewelry, accessories and things like that. And this is the biggest challenge, right? So updating his, so he has physical outlets as well.
And then he also has online presence. This is biggest challenge for him to make sure that each and every product and each and every SKU has good photos to go along with that. the fact that they have on conversion rates is significantly higher if you're able to present the product that's being used in some way, or form. So I think this is a big challenge. And where are you today in terms of solving this problem?
give us some understanding on the status quo as things stand at reshop. Where are we today? Yeah, so we launched last year end of May. our team is currently globally remote. So we are a remote press company and we have teams in Hong Kong, mainland China, and the US as well. And ever since we launched last year in May, we now have over 350,000 merchants using us globally.
predominantly in North America as well as Southeast Asia and North Asia as well. We do have some users from Europe. I would say that's like maybe like fourth or fifth on just our country ranking, but majority of our merchants or users are small medium owners from Shopify, Etsy, Lazada, Shoppe, you know, who really are
Yash Shah (04:31.342)
Basically have a small team and they're trying to have like a ad for that can help them solve the first step into creating model and product for those. is amazing, right? So 350,000 merchants in about one and a quarter years. Yeah, about a year. So that's like close to about 800 merchants a day, is amazing, right? So which is really cool.
This is great, right? it's lovely. so here's what that brings me to as well. In the age of AI, no idea is unique. so because AI also tells me ideas and problem statements and things like that, and it's getting easier and easier to build solutions and build products. And in that space or in that age, how do you
sort of make sure that there's not some competition somewhere out there in the world who's doing a little bit of a better job at the product than you. Like how do you think about competition, right? Because technology is no more an advantage. So like your thoughts on that. Yeah, exactly. And we got asked a question like,
a lot of times, I think from the beginning of our journey. And I would kind of dissect it into twofold. The first is how the macro environment or how the large language models progress will impact us, right? Because we build on all sorts of open source models and that has its pros for us, but also it might impact us negatively. If, know, one day OpenAI came out with something that
just kill all of us small entrepreneurs out there. So from our perspective is from the macro point of view, how do you prevent a mega or a giant company from kind of intruding your area of expertise? And how we did this is because of our know-how in fashion, clothing industry. So really, more to the fashion arena, that is the area that we decided to go even deeper and more niche. So that's why
Yash Shah (06:48.972)
know, 70 % of our users are from the fashion and clothing brands. That translates to how we work on our products and how we strategize our product roadmap as well is that we focus on the AI models phases and how that would be developed into, you know, more natural appearance, very natural generated result instead of like
the very uncanny artificial images you might see from mid-journey. Although mid-journey is obviously great and fantastic, but you know this is from mid-journey. You know this is not like a real person, although she looks perfect and everything. And that is something that we are working really deep into making something that looks natural that you, to be honest, you probably, hopefully, wouldn't be able to tell that, this is AI generated in the future. And then we also work on areas where we
work with actual real model agencies. So we work with a US model agency and a Germany model agency to enact in an attempt that we can train the digital avatar or the digital representations of their models so that we're not trying to kill a model industry, but more so we're trying to merge both roles together so that the traditional models, they're able to find their foothold.
in the new AI role and that really, hopefully would eradicate some of the ethical considerations as well. So that's one area on how we try to compact a more macro view. And, you know, in addition to that, obviously there are like hundreds of smaller competitors out there. They kind of died down in the last six months, I think, for two reasons. One is because of GPU calls are still...
difficult and the fundraising market isn't that optimistic so kind of you can be at a self-sustained age but also because a lot of the images that they generated on for example mid-June wrapper so it just looks uncanny and unreal and you wouldn't be able to get enough e-commerce merchants to accept that and to actually use those on their e-commerce stores or for their you know
Yash Shah (09:08.662)
offline usage. that wouldn't translate to the top line and just kind of, you know, make them go away. So to our perspective is also how we able to stand out among like other majority rapper or companies who maybe focus on just changing backgrounds. And they focus on maybe background removal, like a photo room, that's where their strength is, and probably not ours. And ours is really still within the fashion AI model.
arena. So that's the battlefield that we tried to go deeper and really build our name in that specific niche to start with. Interesting. And so for all the people who are listening or watching us, and I would recommend, think the first point that Lin mentioned about the models at V-shop looking a little more realistic. All of us might have seen
outputs that mid-journey produces. I'd also recommend just going to vshop.ai and just scrolling down. And you'll see a host of photographs that have been generated using vshop and you'll see the difference. And it's difficult for anyone to say it out in words. it's perfectly imperfect when you sort of start to see that, okay, this is a little more natural. And it's not as perfect as mid-journey.
Yeah. And that sort of makes it makes it more natural as well. And so the other piece that I want to understand about vShop and this might be also interesting is that most SaaS companies when they start out their first version of the product, it is OK for it to be quick and dirty. Right. So it's OK for it to be sort of put together. In certain cases, we've seen even no code tools for the first few customers. And it could be a little buggy, it could be a little off.
And things like that. However, in case of V-shop, what my assumption is, correct me I'm wrong, is that for you to close first 10 customers, the product already has to be really good. It is only then that they actually come on board. So your product finesse has to be as good as it should be for a traditional SaaS company at 100 customers for you to acquire your first customer.
Yash Shah (11:32.61)
How did you think about that? How did you go about thinking about it? I think that's a great question. It is also one of the struggles that I might mention in the later part as well. I think for us, when we started last year around the sample and mode, we're still getting early adopters. So they're perfectly fine, if not more, know, better, less vulnerable to imperfect or hallucinated.
parts of AI, so they were able to accept some unstructured limbs or unnecessary peripheral generated with AI or just hallucinated parts like that. So when we actually first our first version out last year, I would say that it's not as bad, like it's still commercially ready. I think that that was the premise. However,
It was a lot more unnatural than the version that we are seeing now. And the backgrounds were also merged less well. And there are some other imperfections that we were seeing last year at this time. And I think the early adopters still took it well, just because they first joined in. And also the user interface wasn't as friendly. The learning curve was even steeper than what we had.
So we were able to get, I wanna say like the first 10,000 are still more so AI enthusiastic or like the really small folks who's like a one man, two man shop that they wouldn't be able to have any images before this. But now they have, they at least have something. It would boost their sales by maybe like, you know, five to 10, not like 10 times, but that's something that they were able to accept. However, when we get to, you know, 350,000 now, now it's getting a little bit harder.
because we're going into the traditional merchants or like, you know, larger corporations who have seniors be like, this isn't perfect. This is not the same work as I would have if I were to hire an agency. So now we're going to that, you know, area where we have to like, you know, check on that area. I think that's the difference now. But you're right that, you know, think that's a struggle we seeing in which now we are getting to everywhere
Yash Shah (13:53.998)
you were expecting a perfect result from all the AI tools, AI software you were seeing, as perfect as if we were to use Zoom or Telnet. But it's just realistic, not the same technology behind it. what a great problem to solve, Where a high-ranking decision-maker in Chanel is comparing an AI-generated output with an agency.
And that's already a high enough benchmark. And this makes me also wonder a little bit where I know for a fact that from the time that you started to today, there would have been different points in time when you would have had to like, typically you start a company as generalists, right? People are like 70 % good at five different
And then over a period of time as and when you scale, the types of challenges that you face are very, very different from a product standpoint as well as from a GTM, like a go-to-market standpoint. At what point did you think about making that transition? Were you thinking about internally growing? Were you thinking about recruiting specialists from outside to solve very specific challenges and problems?
How has that really been of scaling and solving challenges that 90 % of the world wouldn't be solving? To be a little more clear, what I'm talking about is the difference between starting probably a hospital as a group of general physicians and then at some point a patient walks in needing a heart surgery. How do you make that transition?
I think so obviously with a company, there are two kind of teams or departments. can kind of generalize departmentalize a technical and technical. I think a non-technical trip, we haven't seen like a big gap or a big transition from kind of where we were last year this time or where we are now, just because I don't think that the go-to market for AI companies are
Yash Shah (16:18.446)
radically different from what Zoom, what Calendly were doing in the past. It's still the same. You're facing the same group of people just trying to educate in a different way. For technical though, it's radically different because like you said, when we first started out for the first six months, technology was like, it was literally something that's unseen, it's changed on a daily if not hourly basis, right? So for us, we have
maybe one of the best tech journalists group to be able to train themselves up to learn about the most advanced technology development out there. So to learn what Meta's open source is doing, how we're able to incorporate that, we're able to incorporate stable digitalization, all the open source, all what we're seeing from other communities out there. So that's not that challenging to an extent where
As long as you have a smart group of engineers, you're able to kind of do that. And as long as they're fast learners, right? It's not rocket science to begin with. But now we're getting to the, I don't want to say rocket science, but it's like, send my rocket science for the unsolved issues out there. For example, imperfect hints, hallucination and unwanted parts of the AI images.
how are you going to solve that? Because obviously, majority didn't solve it. Stable division hasn't, dialy hasn't. the smartest people on this planet haven't solved it. So now, I guess for us, the transition has become, one, is we could rely on the foundational model to advance, and then we kind of catch up with it and adapt quickly.
one way to do it obviously I think that's an obvious way. We do take that route too however we are trying to find like work around or tricks where we probably invent along the way I don't think like it's you know you can strategize this just kind of like scaling all over it be like you kind of found it and it's the trick that you know we own or we can own a pattern on I think that's the second thing that we now transition into.
Yash Shah (18:39.576)
So that's why, the reason why we moved to remote globally as well as also we're trying to find those talents across the globe. It doesn't have to be like the machine learning engineers who has the stellar backgrounds from top university. It could just be someone who kind of understands the tricks or kind of themselves into it on like a daily basis. And I think that's the area that we are moving into. So like a couple of engineers that you have.
are also scientists. They are right at the frontier of trying to solve a problem that all of us know about but has not yet been completely fixed. That's fascinating. talk to us a little about how did you think about go-to-market? What was the...
What was your first few steps? are a couple of channels that you tried? Some of them that worked, some of them that didn't. A couple of experiments that you ran as a GTM for VShop. Yeah, I love to share with everyone because I think to get to the first 1000 customers who are using it, probably the hardest. I know people go to their friends and families and getting feedbacks and asking for something like that.
didn't like do that. think for us, the interesting thing is when we first started out last year in May, that's when Threats by Meta, it was just a mod. And so I started my Threats account and I started to build in public actually, not on Twitter, not on X or so to speak, because X was already like, you know, full of very successful entrepreneurs. So it's hard for me to kind of start my new presence there and try to build in public and
other AI influencers. that's why I was like, maybe that's what's the way to go. me just travel. And actually that works. So I started to build in public on that and interacted with those also new that's influencers. So they obviously had no exit counts and they were popular or predominant as well. But they were also new on that. I mean, they're not doing a lot better than my account was.
Yash Shah (20:48.238)
I started interacting with them and I got unprecedented reactions where I just wouldn't be able to on X because there was nobody there. And that's how it kind of went viral. then we had, well, I say like viral, viral, we get like, know, accounts that feature us and they got like 200,000 views and there's some comments and everything. that's when the first inflow or the first flux came in. And then we got to, I want to say like,
50,000 quite quickly just because everyone was looking at AI and people were trying out new AI products. I'm sure you were as well trying out like I myself tried everything. I don't necessarily pay for them but I would sign up for a lot of things and to see premium or free trial and I think that's how we got to that number quite quickly. So I would say like for the first early adopter journey,
luck and hard work at the same time. Like you kind of have to find the right place for you to fit in. And I think for us, that was the right place and the right time. then moving on to that though, unlike most SaaS companies, which I'm sure most people probably would, they launched on Producton. And do you think Producton is a great platform? If your product is like a notion, right? You cater to SaaS founders or like business entrepreneurs.
For us, because we cater to e-commerce merchants or fashion retailers, they don't really go on Product Hunt, to be honest. They go on like Shopping by App Store or other e-commerce platforms to shop for their app or their necessities. That's why we actually never done Product Hunt. Plus, it's a lot of work. So I think it is important to consider where your users are and go there and set up. For example, when we did threats,
Well, I knew AI enthusiasts were there or on X and I chose the format just because it was easier to tackle and we went there. And as we moved to kind of the second stage of our go-to-market, I knew like Shopify apps are somewhere that we should be in. And then that's why we chose to get listed there as well. And now we are ranked number two among the image editor or the AI e-commerce image editor category.
Yash Shah (23:11.016)
We're working on the typical list, so we're trying to move on using SEO and a lot of users' reviews and everything. I think the key is to go where your users are. I don't think product support for anyone is really more a vanity metric than, you know, just need to think about where your users are. Yeah. Yeah. So I've launched on ProductHunt about
about 13 times and I can very confidently say that if you want criticism, you can go to Productant. But if you want users, if you're users, you can go somewhere. So I think Productant is great if you're looking for feedback just from your peer group. They are not your customers, but from your peer group, if you want feedback, that's probably helpful. yeah, so customers are...
mean, having 13 launches, a couple of them for our own products. I think about 10 or 11 of them for our clients on behalf of our clients. We're building their SaaS products. so, yeah, we see a bump and then it sort of goes away. Interesting. So would you qualify these two things? think you mentioned luck and I want to...
qualify that a little further by saying that I would not, I wouldn't call it as luck as much as I would call it as headwinds in the market. Yeah, I agree. So because like everyone is talking about it, you want like the market is also looking for it and so it's always helpful, right? Yeah, it's about the right time. It's called headwinds or tailwinds.
But yeah, whatever is helping you go forward. am forgetting the analogy. Yeah. Probably tailwinds. Yeah. Is it tailwinds? Yeah. think so. So tailwinds is something that goes opposite. Yeah. They're pushing forwards. Yeah. got it. So then market tailwinds, right? So, yeah. Got it. So now I remember this because I made this very public embarrassing mistake. So, so market tailwinds, right? That push you forward.
Yash Shah (25:26.478)
And then second would be would be like marketplaces where your users already exist. Would you be open to talking about a couple of mistakes or not mistakes necessarily, but a couple of experiments that didn't work on a GDM site? Yeah. So we did, I think the first thing is we did meta marketing or like meta ads, like most people would, you just run instant ads.
Facebook as we're done knowing why you're doing that, but you're like, but everyone was doing that. My manager came along and be like, you know, this and that. Your competitors are doing what, what CPM, what CPM. We're like, okay, let's do it. And I think that was a mistake because so we didn't have a mobile version back then. We are developing our app and it's launching new version. So you guys can check.
that afterwards, but we still don't yet have like a sophisticated mobile version just because like if you look at major anything, they don't either just harder to kind of play around with everything on your mobile and cars with different skills. And we still ran Instagram ads, which now in hindsight, why are we doing that? Because people only look at Instagram on their phone and they will be directly to a landing page where they are then asked to
go on their browser to, you know, enter the site and to start working on it. was just such a long funnel that wouldn't work. And it was very hard to measure the actual ROI either. for us, so we didn't spend a lot of money on it. It was more like a trial and error and it proved to be more error than success. Obviously, I'm not discouraging running ads, obviously, but I think like as
We were opposed to product market fit already when we started to do paid marketing on a very small scale. But if you were a pre-product market fit company, like you don't really have revenue, like $10,000, I wouldn't recommend running less just because first of all, it could go really out of your game. Like it could go to the booth without controlling a budget. And the second is the users that you get is hard to
Yash Shah (27:49.366)
to gauge like the actual interest and everything. It wouldn't be measured as PMF. And I think the second thing is something that we haven't done. I think fortunately or not, we didn't do any SCM. So we do a lot of SEO, obviously. We focus on long tail keywords, our image SEO, as well as content SEO. But we haven't done any SCM. And the reason is also
It's still harder to measure the ROI just because going from the top funnel from your AdWords to user coming to your site to registration to let them pay. I think for our scale, it's just harder to justify the cost at this point, given that there's still a lot we could do with SEO. So I do think that's something that we...
we haven't done and I think working on your SEO was better than just buying at words at this point. And we actually saw competitors buying our brand name as like they just buy WeShop AI, right? They just like spend money on it. And we still ranked number one regardless just because Google is still quite, know, I want to say like clean to the extent. So yeah, so I think like SCM, you wouldn't be able to do that.
for like a decade. Karan, it's a couple of quick things. One is when a competitor buys your keyword, it's a proud moment, but it's also a terrifying feeling because you don't know how much money they have and you just don't know, right? So that's one. And I tell you what kills companies with meta ads. And in my very limited experience, one of the things that I've seen is that running meta ads, it doesn't fail completely.
And so what happens is, is that it works just enough to give you hope. Yeah, exactly. That week will be better and next month will be better. Maybe I'm not investing this and maybe I'm not doing that. And so like if it doesn't work, like it gives zero results, it's a very easy decision, but it gives you some result where you think, okay, instead of thinking about whether I, whether this is meaningful or not, you start thinking about how can I optimize the results? Yeah. And I think that's
Yash Shah (30:07.242)
That's what leads to a lot of wastage in terms of dollars and hours of effort as well. So thank you, for the time. This brings us to the last part of the conversation. I know we are going above our time, but I think this is helpful. Just a couple of minutes more. So first thing is, I have a question for you that, you know, previous guest who was David, David runs a company called DataBug.
and data bar basically offers you ability to pull in data from thousands of different applications that you can integrate into it and then offers you a spreadsheet like interface and you can create charts graphs and so on and so One of the challenges that he's trying to solve for is the conflict between what are the things that he wants to build that he believes are
good for the customer versus what the customer is asking for today. So there are things that as an entrepreneur like you also must face when you think that the customer is not asking for this because they don't know something like this is possible. So the customer is always asking for faster horses, but what if you want to a car? And so how do you at Vshop sort of try and maintain the balance? How do you think?
That's a great question. For us, we do encounter similar things. So for example, as I mentioned multiple times, everyone knows that AI generated this form of limbs or this form of bones. It's not perfect. But they will always ask like, is there something I can write to make it better?
Or is there like a tool that we can do that without knowing? From a technology point of view, you can fine tune something that was able to kind of sort of rectify that mistake or that issue. Although it's not perfect yet. So I think for us, it was more so instead of asking what the users want, like I'm sure they want
Yash Shah (32:26.104)
I mean, what they want is a perfect hand, right? But they would ask a question or they would phrase a question, be like, what problem I should be writing? Or are there any tools that you recommend or you should build a tool that can help with that? But all they want are just perfect limbs or it doesn't have to even show their hands. It's not like you need hands to showcase your dress, right?
So there are tricks around it, for example, if you hold a fist or, you know, if you just kind of tie your hand in pockets, you're to generate something. And that's like the easiest way instead of trying to, you know, write like thousands of, you know, prams over it. So I think it's more so, so we never asked what do users want? It's more like, what issues are you facing? Let us know your, what your image is so like, the before and after. And we would go into it and to see, you know,
what could change though, what could correct those issues at this point. But obviously there's still a lot of things that are just, you with our limited technology, we are unable to change all that at this point. But I think instead of asking like, what do you want to say? What issues are you facing? Yeah. Got it. Interesting. And what's your challenge? What is it that you are trying to solve for today?
Yeah, so I should ask my next customer. Yeah, like, I think like hands are definitely something. And then I think for the next guess, the next guess is also in the AI industry, which I think a lot of entrepreneurs probably are. Yeah, yeah. I think with AI, because I just read something new today where cloud, cloud AI, so Anthrobics, they just hit 1 million ARR.
Well, they have only right that that excludes their API and everything but imagine like that's Literally the number two AI company out there Behind opening and media is in its own. Yeah, that's different and their app has only hit 1 million AR within months of launch and That just basically says well first of all people aren't necessarily paying for AI as long as you know, you have some credits
Yash Shah (34:49.774)
I can tolerate that and I can kind of, you know, use the limited features out there. And the second thing I think it's, for example, with us, we know the main issues of our pain rate or why people aren't paying is because it's not perfect. If it's as perfect as you have, you know, this black shirt of Yash and you just take a picture of that and boom, there you go and handsome Indian man.
muscular dressing that I'm sure you would buy the software immediately, you know, in a heartbeat, but that's just not how tag is operating right now. And we do hope that wouldn't appear one day, although it might disrupt our current landscape a lot, but because the current AI or what it could do, isn't really a hundred times your efficiency or your productivity, or it's not something that you envision in your head on what AI would look like.
And that's why people try a couple of times and they'd be like, this is too difficult. Sit button curve, I'm not pulling for this. I might use it or I might revisit it in another time. I think the key issue for us is how do you kind of compact that with the current AI landscape is growing. It's now, think, a plateau over the past three months, unlike, you know, a year ago. And you will probably see some companies revenue hitting a plateau as well. So how do we kind of all face that?
doing to wait for HBD5 or something to appear. If not, then how do you as a bootstrap founder tackle this? That is a brilliant question and I'm surprised. I'm not surprised. I mean, I just feel sad. Why haven't I thought about this question myself? That's great question. wouldn't be able to answer that. That's literally like, I don't know how. Yeah. Yeah. No, this is...
This is good. And this is good, right? Because I love to see my guests sort of find it extremely difficult to answer. And what I found, like we released about eight episodes of the podcast recorded, think, and after 20 episodes, after recording 20 episodes, one of things that I've realized is that the guest actually asks the more difficult question, because it's something that they're going through. And so this is helpful.
Yash Shah (37:14.85)
But Lin, thank you for taking out the time today and patiently being part of this conversation and for all the people who joined us on Shopify or on YouTube or on Amazon Music. Thank you again for viewing or listening to this. you go into the comments, you'll be able to find the link. It's fairly simple. It's vshop.ai, but you'll also be able to click on that link and go to their website as well.
own an e-commerce store or work at an e-commerce company, you give it a try. It's a platform that will genuinely bring 10x more efficiency in your product photo shoots workflow. Thanks, Leem again. Yeah. Thank you, Yash.
.png)
.webp)
