William Leonard

Welcome back to the Atlanta Startup Podcast. I’m William Leonard, your host, and investor with Valor Ventures, a leading seed-stage VC firm here in Atlanta, Georgia. Today, I’m excited to welcome to Podcast, Atlanta entrepreneur Tim Hall, co-founder and CEO of Simporter. Tim, welcome, and thanks for joining me today.

Timothy Hall

William, thank you very much for having me. We’re really excited to be here. 

William Leonard

It’s great to have you. For the benefit of our listeners, we would love to have you provide a high-level overview of Simporter and what you’re building?

Timothy Hall

Thanks for that. Simporter is an AI company. The whole mission of it is to help larger consumer product companies predict if their new products are going to be successful. We use AI in a really novel way, where we’re examining really large pieces of data and really big amounts of signals from consumers but then trying to bring that all down and distill it down to help a big product company make a go/no go kind of decision on bringing a new product to market. 

William Leonard

Taking a step back here, Tim, I know you’re local to the metro Atlanta area and you’ve started entrepreneurial endeavors in the past, but we would love to give our listeners your journey to starting Simporter.

Timothy Hall

I’m probably older than most of the founders you would talk to. This is my third startup. Before I went into the startup world, I grew up in large public companies. I started my career with Procter and Gamble, Cincinnati, worked for Hasbro Toys, and I came to Atlanta in ‘99, to work for Turner networks. I always had that sort of big company experience. But most of the engagements I had were product companies where I was putting stuff out into the marketplace, coming up with new ideas with the team, developing them, and then seeing how others sell at Walmarts, Best Buys, and Target stores and places like that. My first startup was a consumer products company called Digital Blue. We had about 110 employees in Atlanta at the time, this was 2002-2012. But we had the same problem there that we had at Procter and Gamble and other companies, which was there are no magic eight balls, no crystal ball will tell you if your new product is going to be successful. But the market got a lot choppier and faster over time with the introduction of Amazon and other eCommerce players and with this big pivot to peer-to-peer marketing and peer-to-peer conversations. In the old days, we could run TV ads, we could run coupons, and we could really drive consumer behavior to try new products. If you have big budgets, it was relatively easy. If you have small budgets, you have to be more guerrilla about it but you could really move the needle, sort of talking down to the consumer. Now it’s all peer to peer where it doesn’t matter how big your budgets are, if your product’s no good or they don’t really need it, they won’t try it because they’re reading reviews, reading what people are posting about your brand and social media. You have all these little nimble startups that can enter the market really easily thanks to all these great new technologies like Amazon and Walmart that let you sell products on their platform. Shopify lets merchants have a really good store environment. Instagram you get this really great viral marketing at scale. If you’re a traditional brand, there’s just so much going on in the market. It’s so choppy and frothy that it can be really, really hard to predict if your new products are gonna be successful. I lived that journey for a couple of decades. It got tougher and tougher to get a new market to succeed. Statistically, about 9 ½ out of 10 new launches fail within two years. It’s a tough marketplace. We saw the opportunity to bring some quarters of technology to bear to try to fix that problem for all kinds of companies.

William Leonard

That makes sense. As you thought about starting the business, kind of talk to me about the pain points of how retailers and manufacturers truly went about determining which products consumers wanted prior to these AI-based solutions really began to enter the fray.

Timothy Hall

On the brand side, the manufacturers usually would kind of look backward at what they had launched before. You’ve come up with a new product idea, and it kind of passes some gates internally. There may be a little bit of consumer engagement about the new idea, some qualitative testing like focus groups, and there could be some panel research to get a couple of 100 consumers to read and tell you if they would try it or not in a test environment. But many, many times, we think this new product, or this new line extension, or this new brand is going to look like these old brands that we launched a year or two ago. They look back at that sort of small set of data and say, “We kind of think it would do this many units.” You might get a series of opinions, where they might go talk to the sales force or talk to the retail buyers. This combination of a little tiny bit of data, some analog products from the past, and then a bunch of people’s opinions would all come together to decide if we’re gonna make this multi-million dollar investment. It’s just not really good science going into those predictions. That’s how the brands really do it. The retailers have a different position. Every manufacturer comes to him and tells him that their product is absolutely the best. The retailers have to, first of all, lower their risk, or de-risk that as much as they can. One way they like to do it now is to try it online first and there’s not much risk to a retailer if it’s online on their e-commerce shops, or we’ve even seen some of the big mass merchants say to the manufacturer, “Go try it on Amazon first and then come to us with the results.” I actually had a merchant tell that to me back in 2012 saying, “I think your idea is interesting. Let’s see how that performs on Amazon and come back to me with the data. I’ll decide what I put in my brick-and-mortar stores.” From their standpoint, as a retailer, you’re presented with hundreds of new concepts a week, and your job is to evaluate them, figure out which ones you’re going to spend your very valuable space and resources, and time on and then figure out how to de-risk that opportunity as much as you can. If it’s one of that 95% and fails, it doesn’t cost you too much-lost sales opportunity or lost margin. They approach it differently between the brands and the retailers but they’re both stumbling against the same problem, which is not great data out there to tell you prior to these AI solutions. There haven’t been much great data out there to tell you what’s going to work.

William Leonard

I think that that overview really paints the picture well as to how things were done in the past and where the efficiencies really inserted themselves. Is it accurate in saying that the core mission of Simporter solution is to predict product demand prior to and post-launch? Is that correct?

Timothy Hall

Exactly. Accurately predict it.

William Leonard

Got it. I think our listeners would wonder what is the practical roadmap to truly predicting demand for new market concepts? Could you elaborate on the factors that are helping ensure the accuracy of these insights as well?

Timothy Hall

Sure. The way our model works is we start by ingesting really large pieces of data. First of all, we like to look at a category. Let’s say you want to launch something in a food category. We’re global partners with Nielsen company, with Nielsen IQ. We have access to their historical sales information for that category from all the cash registers in the market. It’s the United States. We can start with a foundation of three or four years of historical sales. Now, this new product, that’s our hypothesis here is, it’s never launched anymore, so we don’t have any sales for it. But that new product idea is composed of lots of little attributes. Those attributes could be flavors, ingredients, the size of the package, those attributes could be its price relative to the competition. Some of those attributes are more ephemeral, like what need does it solve for consumers, what we call need states. When we know what that new product is, it’s composed of all these different attributes, we can go back over our historical database and look at how those attributes are done at scale in the past. For every product that shares one of those attributes, glom them all together and see how those do. That’s kind of the baseline. We use that as training data for these other integers that we bring into play. The other stuff is what our consumers are saying today. We pull data out of eCommerce sites from Amazon, and Walmart, and all these other different eCommerce sites for all kinds of products that have similar attributes. We can see what consumers are writing about in their product reviews about those attributes and those specific individual things. Then we pull data out of the search engines and say, “Are people searching for those attributes? Have those searches increased? Decreased? Are they flat?” Finally, we pull data out of public posts and social media. When we’re talking about attributes what are people saying about it, on Twitter, Instagram, Facebook, etc. All protecting privacy, it’s all anonymous data. Now, we’ve got this huge lake of data on all these attributes that go into your new product. We can effectively synthesize that product, or create a synthetic version of it in our data stream, to say, “Alright, this product assembles all of these things including from the more hardcore stuff like price and distribution down to the more ephemeral stuff like needs that it solves.” We can get a pretty fair, accurate benchmark of how it’s going to sell in the distribution footprint that the brand or retailer thinks it’s going to be in. The way we measure this, your other question is, how are they accurate, on average, for all the projects we’ve done over the last two years, we’ve had 86% accuracy. If we say, it’ll sell 100 units a month, we’re right to 86% of the time. We measure that by going and looking at what actually sells so when we do these predictions, and we say that new products are going to sell x after it launches, we track it, get the data, of course, and we validate that we were right, and 86% of the time, we’re right on the money.

William Leonard

Wow, that’s a very impressive hit rate there. Kind of going off on the accuracy tangent, how are these AI algorithms evolving with the enterprise and/or consumer buying behaviors and equally along with the macroeconomic events that are happening around us?

Timothy Hall

The AI models that we employ are proprietary, but they’re not ridiculously complicated. We’re looking at really large amounts of data and running calculations on them. But it’s not as challenging as some other areas of big data. Because these are all sort of what we would call cause and effect kind of integers. The hardest part about what we’re doing in these models is to make sure the integrity of the data coming in is good. It’s what we call cleaning the data to make sure there’s not a lot of noise or disruptive stuff in the data that’s going to show the results. But the beauty of artificial intelligence is it learns and corrects itself over time. We can get that accuracy because when we have three or four years of training data, our AI model basically has a hypothesis of all that stuff I talked about social search, etc. All that stuff happened to some degree in 2019 on a certain product. The AI model says, “Well, here’s what my prediction was for that 2019 product.” Then that prediction is wrong. Because it’s still in the early stages of training, and it can see what the actual sales are, then it corrects itself. When you think about that corrective aspect, that’s machine learning, that’s why the models become so accurate, because they go back and look at large data sets, train on those datasets, and get better with all those cycles of corrections. When we work with a client, and that product launches to market, it continues to self-correct. That’s why AI models are so valuable compared to other ways of doing it because it automatically does that correction without humans having to be involved.

William Leonard

Right. That makes sense. Kind of looking from the customer perspective, here walk me through the customer experience from the first meeting or discovery call to the implementation of the software in the models, and what does this process look like? Is there a typical timeline according to you know, a varying customer profile?

Timothy Hall

There are two different kinds of clients or customers, two different kinds of personas for us. A very typical one is somebody who’s been tasked at a corporation, with developing new innovations. They’re trying to find something new and different and unique to bring to market so they’re not involved in the nuts and bolts of products that are in the channel right now. But they’re trying to come up with the next big thing. Those are usually called brand innovation teams, or they can be marketing personnel. In many, many cases, they’re called consumer insights, the old market research teams, the consumer insights teams now. Those people come in, they’re used to using traditional tools that you might get from other research companies, like our partners, Nielsen, or Acupoll, or GSK, are many of these different kinds of research outfits that do quantitative and qualitative testing. One of the biggest differences is our product or software and there used to buy ad hoc reports. We hired a team of analysts on our side, to run the software for them and the first engagement and prepare a report because that’s the currency that we’re used to working with. Once they start getting reports from us, they get more interested in how these dashboards of our software work. They say, “You mean I can check this anytime I want?” And we say, “Exactly.” We have customer success people who kind of begin to show them how the software can be used day-to-day. But most of our clients need the product during certain periods of time when they’re working on the new products or their different sort of development paths and development gates. We try to be really responsive to those. Because sometimes they have really hard deadlines that come up, and they need data faster than others. Your second part of your question is, is there any typical engagement profile? The answer is not really, because it depends where in their business cycle they’ve been introduced to us when we’ve come in. We’ve had clients that have seen what we do, and they want to start the next day they say, “We’ve got two weeks to get to validate these products.” We had one company that we worked with last fall, I can’t say the name, but they were in the personal care space. They had a new chief marketing officer join and a new team came in and a new CEO. They needed to get an entire product line across two different big categories to market within 12 months, which means they had to show that product line to their retailers by January, and I think we met them in November. They had their own design concepts and had us run all those concepts. I think we ran through over 80 different products through our solution to help them pick the winners. Sometimes it’s like that, most of the time we’re working on something we’re going to launch in 18 months, and it’s usually three or four months to sort of see how the solution works. they usually move into a 12 month ‘Always On’ contract where they’re accessing the software all the time.

William Leonard

Got it. Got it. You kind of spoke to it in that previous answer. But what are some of the benefits that your customers have seen and have really been able to quantify by using Simporter?

Timothy Hall

I think one of the biggest ones is gross margin growth. There are two ways that they increase the gross margin for a product: one is a big component of our application that tells them the right price to be at because the price is an attribute that we examine. We had a really big client, a really well-known food company that was operating with us in Brazil. Brazil has a really tough economic headwind right now. COVID plus it’s just a tough market. They had some new product introductions. They first use us to validate the product introductions. They had certain benefits that made sense in today’s market. They wanted to validate those that were going to do well in the market. We validated and then we said, “Your price could be 25% higher.” They were skeptical at first. We said, “Here’s the data.” Because price sensitivity and price modeling are all part of what we do. They say, “Well, we couldn’t possibly raise the price of that particular segment because it’s tied to another segment in the marketplace, the same size containers. We couldn’t bring both of them up. We couldn’t bring them up separately.” So we said, “Well, let’s just run the analysis. We’ll enter these other products that you’re thinking about that are already in the marketplace so they could have upward price mobility.” They did, so this client was able to take the price up 25%. Despite the economic headwinds, it’s doing quite well. Basically, it was like a single surf product and the consumer was pretty flexible, whether it was one price or the other. The data got them a much higher gross margin. So that’s really the quickest way for them to determine if they can have a higher gross margin by that price validation that we do. The second side of it is by just getting the products right. When we’re using our AI solution, it’s generally predictive. We will tell you, “Hey, this product is going to do X units per month over no x period of time. Here’s the life cycle.” But another piece of it is prescriptive. We say, “These are the attributes that are leading it.” We’ve had clients kind of change their messaging a little bit because our analysis shows the uptake by the consumers is stronger on certain attributes than others. In that case, they get higher sales, it’s harder to prove it but price examples are really easy to prove that we directly generate the margin. But in this case, there has to be this qualitative feeling among the client that repositioning is based on the data they got from us that generated the higher sales. But there’s real validity to that because you know, a few tweaks to the package or a few tweaks of messaging can make a big difference with consumers who have nanoseconds to decide whether to try something in the store. We generate greater sales from that side of it as well. Those are the two main ways we help a client. The third way is marketing investment. When we show them the products not going to be successful, or if it’s in the market, we help the big food company in Europe that was launching a product that had done really well in Western Europe, they bring it to Poland, we show that it was not going to be successful in Poland, even though they’ve been successful in other markets. They didn’t believe it at first and they believed this, and they pulled their advertising off the TV off the air. We were able to save a lot of marketing money because they got an early warning from us that it wasn’t going to work well. We don’t like having those conversations. But you know, the data is data, and at least they knew to be able to kind of reduce their risk of loss and protect their margins there.

William Leonard

I mean, that’s super impressive that you’re able to help these companies on that scale, in such a proactive fashion. Kind of pivoting here a little bit to more about the team of Simporter, many of our listeners may or may not know this, but you and your son Dillon are actually running this business. Can you talk to us about that dynamic there and how it’s really allowed you all to grow this company?

Timothy Hall

I wish we were on video because my joke makes more sense there. Because I always say, I’m the beauty, he’s the brains. I have to assure your listeners of beauty, but he’s a smart guy. I’m lucky to have a co-founder, who is my son, because I kind of really came to our business understanding the problem side of things really, really well, because I lived it, launching products that didn’t succeed in the marketplace. He came at it from the solution side. He was in Europe, he’s still finishing University. He was working in Europe for this lab that was funded by Google. This is in Poland of all places. But he was working on machine learning models for this Google-funded outfit. I was telling him all about the challenges because I was considering launching yet another production company back in 2018. Because I’ve been in product companies for most of my career, I was explaining the risks of forecasting badly and launching bad products and stuff. He was telling me about what machine learning does with large data sets and if you could find the data out there, you could automate the answers and have a much better outcome than the old way of forecasting. We’re literally drinking beers. He was old enough at the time. He’s 21. We were drinking beers in Warsaw, talking about the problem and he was telling me about the solution. I didn’t know anything about artificial intelligence machine learning other than Alexa, and we decided to build some software which, at the time, I was considering just like something I would use to validate products that I would go build and bring to market for my next product company. We started working on that together, and it was really great working with my son, it still is, because he’s one of the few people who can tell me and hold it to my face without me getting my nose bent out of shape. He’s really, really quick to remind me to stay down to earth and to keep an open mind. I think because I’ve been in business so many years and you have these blinders on about, well, it’s the way it worked back in 1995. He reminds me he was born around then, the market has moved on. It’s a really great partnership. The other thing that’s really valuable about this partnership is I’m in Atlanta, where I feel like I’m closest to our market and their customers. Of course, our customers are in Europe as well but most of them are in North America or Latin America. He is based in Ukraine, and he just likes living out in that part of the world. The lifestyle is terrific. It’s a wonderful, wonderful city. Our development team is there. As we started this company, we began hiring developers that we could afford. It’s tough to afford developers in Atlanta because there’s so much competition. We put our development team in Kyiv because they have tremendous talent at a really, really good cost structure. We started out only with the development team, the software engineers, and then we were able to get these data scientists with really terrific experiences that would be beyond the ability of a startup in the states to afford unless that startup got to B round. But we have them on board, we had them on board before seed due to that cost structure. We’ve got 32-33 people in Kyiv, who are really, really strong. I just came back this weekend, after spending a couple of weeks there, just some rockstar talent that my co-founder works with every day. He really kind of minds that aspect of it, overseeing the development piece of it, and I stay more focused on the market side.

William Leonard

I think that’s a very good idea that you all are playing to your strengths, respectively, and unique that Dillon is still halfway across the globe, and you all are still building this successful business. Speaks to the dynamic that you all have as father and son. That’s awesome. Looking here, as we wrap up this episode, Tim, in the present day, March 2021, from your viewpoint, what is the current state of AI? How do you see it evolving over this decade?

Timothy Hall

It’s a great question. AI is probably in the walking stage, it’s been crawling now, I think it’s walking. I’m using an infant example. It’s a walker, not a runner yet. What is really fascinating about AI is we tend to think of it with really, really complex tasks, we tend to think of AI as being useful for massive undertakings. But what we all use it is for some really mundane stuff like Alexa and that’s how we find out the weather in the morning, this is how I do it or Siri, and when there’s going to be interesting about AI as it develops, is it is foundationally changing a lot of stuff, particularly career paths. I think that’s going to be a huge issue in the future because so much of what we would call a white-collar workforce is going to find large pieces of their work being able to be done through AI models accurately, efficiently, and so forth. We’re all going to have to adapt and figure out what the economy looks like when so much work is going to be automatable. You know, like all desk work. That’s a big issue right now in front of us, but what I think is also really interesting about AI is, it’s going to be very useful for little things that you never even thought of, and we’re gonna see AI not just on big massive scale things that huge pioneers are doing and huge unicorns are doing, but we’re gonna see models using it in some really smaller, interesting ways, in our everyday lives. That, to me, is just fascinating and I have a front-row seat too.

William Leonard

I think it’s definitely gonna become a leading factor in our day-to-day over this decade for sure. Tim, this was a great conversation, man. I think our listeners will find a lot of value and insights that you shared relevant to Simporter and the AI market as a whole. I’m really excited to see how you and Dillon and the team continue to build out the business here. Very appreciative of you joining me here today. Let’s keep in touch man, as you all continue to build.

Timothy Hall

Thanks, William, I really appreciate that. I want your listeners to know that we’re simporter.com, it’s really easy to find us. I’m happy to take any questions that people may want to send our way after your show. But I really want to thank you for having me on. It was a lot of fun, and I really appreciate it.

William Leonard

Tim, have a great day, man. Bye.

Timothy Hall

You too. Bye, bye.

Lisa

Thank you for listening to the Atlanta Startup Podcast. You know, we’re not just a podcast, we’re a community, and we’d love to see you at one of our digital or physical events, go to valor.VC and sign up for an event that makes sense for you. We have events for founders and the investors who back them. Another event you might enjoy is Startup Runway. The Startup Runway Foundation is a Valor organization that provides $10,000 grants to founders who are women or people of color building next-generation software products. Applications are free and we’d love to hear from you at startuprunway.org. And as always, thank you so much to the organizations that make this podcast possible. Not only Valor Ventures, but also Write2Market, a tech marketing and PR agency in Atlanta, Georgia, and the Startup Runway Foundation and Atlanta Tech Park Valley’s headquarters, and also headquarters for over 100 local entrepreneurs, building global businesses. See you next week. Please bookmark the podcast and join us.