Product Placement Is Key To Sales And Profit in Supermarkets and grocery stores..
Every aisle of a supermarket has 100s of different brands of products, all vying for your attention and aiming to ultimately end up in your shopping cart.
Getting your product into a new store is hard enough on its own. How do you measure your product availability, inventory levels, product placement or even know what your competition is doing in a store that you do not own?
This data is hugely important for CPG (Consumer Packaged Goods) companies, but has been largely unavailable before.
Our guest today on THC, is aiming to solve this problem with cutting edge image recognition tech!
Today, we have Erik Chelstad, Co-founder and CTO of Observa, a technology company that helps CPG brands understand their customers better through sophisticated data gathering at the brick and mortar stores, using AI to collect data from the pictures, and delivering rich insights to the brand owners.
Their image recognition technology can increase sales force productivity, improve shelf condition insights and help drive incremental sales.
Jed Tabernero [00:00:02] Have you ever gone to a grocery store and had trouble picking the right brand for something as simple as mayonnaise or ketchup? It's really hard for consumer brands to differentiate themselves in grocery stores. So brands often pay a premium for great shelving spots or beautiful displays.
Erik Chelstad [00:00:21] There's a customer we're working with Kona Brewing. They're based out of Hawaii. And they were the beer of the month for a chain of stores across California and also met these really beautiful displays that their marketing team had built up. So they made cardboard cutouts and they had fold like real world surfboards up there. So this huge display stack of beers. I mean, this is part of the customer experience, right? This is one of the reasons you like to go to stores. So we do as people. It's a beautiful display. And and maybe you're going to try this beer. What we do is we went out and we found out the first weekend only 20 percent of the stores had actually set up these big, beautiful displays.
Jed Tabernero [00:00:59] That was Eric Chelstad talking about how his company observer helped Kona Brewery find out that their products weren't getting displayed properly.
Erik Chelstad [00:01:09] They knew exactly the 80 percent of the stores they needed to contact and are working with their distributors, et cetera, to get these things built out. That also led the marketing team to understand potentially why their giant investment in all these displays did not have a marketing. You did not have a sales impact that they wanted it to.
Jed Tabernero [00:01:27] Eric is the co-founder and chief technology officer of Observa a company that is filling an information gap in the retail market for CPG brands.
Erik Chelstad [00:01:37] They want to know what's going on, on the shelves and the stores. They might have a goal in mind. They might be trying to renegotiate with the retailer buyers to get more shelf space for their new flavors, for a bigger jar or whatever.
Jed Tabernero [00:01:49] Observer hosts a team of over 300,000 observers combined with their ML technology, to help provide brands with the visibility that they lack in brick and mortar stores.
Erik Chelstad [00:02:01] We're going to basically create these opportunities, is what we call them. We put those into our system when it goes out. All of our observers that are anywhere near that are going to get notified of this. And we can actually do more than just do an audit of what's happening on the shelf. We can actually fix problems.
Jed Tabernero [00:02:16] If you'd like to learn more about how service is finding out how brands are eating the shelf space, stick around.
Erik Chelstad [00:02:24] We are making it so retailers and consumers, consumer packaged goods brands are able to make real time decisions and and provide a really solid customer experience.
Jed Tabernero [00:02:44] Welcome to THC, where we unpack the ever changing technology economy
Adrian Grobelny [00:02:49] hangout with Jed, Shikher, and Adrian as we tackle the industries of tomorrow.
Shikher Bhandary [00:02:55] This is things have changed. The story is about how Taichi Ono was like the legendary founder of Toyota in the 70s, Toyota wasn't a big company, it was just a parts producer. And he wanted to crack into the US market, the world market. But he wanted to do it in a way that's not as wasteful because Ford and GM were just cranking out cars and just throwing things out there. Let's throw whatever on the wall and let's see what sticks, that kind of model. Right. And they were suffering from it. They were experiencing losses, margins, all that stuff. So Toyota owner visits an American supermarket, right? It opens the door for the fridge for the milk section in the milk section, picks out a milk carton and. What happens is the next milk cart just slides down the rack, right? Anyone who's been in a grocery store knows about this, right? You take one milk carton and the next one slides down. Now, this was a novel feature in the 70s. He was so fascinated by this idea that a consumer can walk in to a grocery store, take what is what he wants, take what only take what is needed. And the supplier stocks only, what is consumed now, this might seem like, you know, yeah, obviously you would make only what the consumer wants. Today, we live in the world of, like Amazon and stuff where the notion that the consumer is always right is common. Back then, it wasn't right. So this kind of ideology about reducing waste, focus on inventory. Kind of led it's called the supermarket method or just in time method are commonly now called lean manufacturing and it's basically what took Toyota from just a parts producer to the biggest automobile manufacturer in the world. And every organization uses this is lean with regards to consumers, employees, whatever. Right. So it's so fascinating that today we're kind of diving into brick and mortar supermarkets and trying to understand what the operational needs are and the data that they are starting to leverage to get better insights with what's selling, what the consumer wants and so on and so forth. So today, we're really excited to have Eric from Observer, a company that helps brands CPG, largely consumer packaged goods companies understand their customers a bit more. So they work in the field of data and they are acquiring that from your retail grocery stores and giving that data. Providing actionable insights to the owners of the brands themselves, so it's great to have you on, Erik. We love this organization. We love this topic of operations and supply chains. And I don't know if you've noticed, but we have a lot on that. So, yeah, it's just great to have you on.
Erik Chelstad [00:06:12] Thank you. Yeah, it's great to be here. Thanks for that entero.
Shikher Bhandary [00:06:16] Yeah, no, absolutely. And so right away, I kind of wanted to dove into what problem you guys were looking at because you were looking at brick and mortar stores specifically. So if you kind of can give us the lowdown as to how what problem you guys are looking at and whether you like what made you decide to actually get into the space because there might have been a trigger of some sort. Right.
Erik Chelstad [00:06:46] Yeah. Do you want do you want the real human version of this? Yeah. Is that another one would love to give. Yes. There's the one you give to investors about how you analyze a certain space in real life market opportunity. But I'm going to tell you, I would like
Shikher Bhandary [00:07:00] this synergy and stuff like that, OK?
Erik Chelstad [00:07:04] Yeah, but I mean, it's like it's more interesting to know that my business partner and I were probably having a beer and actually talking about beer. It it came around. So I, I my background is engineering and technology person. I also I grew up in the Pacific Northwest here and I spent a lot of time growing up in Alaska. And so I love being in the outdoors. And it may seem weird, but I like combining those two things. I it's kind of it's real because you can become a computer person with all of your downtime. You can study and encode and write things and then go outside the few times that it's actually sunny and nice and try to go climbing or skiing. And I was volunteering with an organization here called the Northwest Avalanche Center and trying to figure out ways to get information from the backcountry. And so the backcountry is populated largely if there's anybody out there at all. That is talk about the mountainous backcountry where you have to hike. You know, you're you're hiking for two hours on a pair of backcountry skis to get out someplace or you're riding in a snowmobile or whatever you're getting out there and you do not want to die in an avalanche and you don't want anyone else to either. So working with NOAA, we were trying to figure out how can we get people to give us information when they're out? And so I built an app that was basically something that could be used by people in the back country to relay information back to the these forecasters there, like avalanche forecasters, are like weather forecasters, but they also do talk about avalanche potential
Shikher Bhandary [00:08:47] just bootstrapping your way.
Erik Chelstad [00:08:48] Yeah, well, it's a nonprofit, right? So I'm like, how do we do this? Can we hire somebody now? Let's have Eric do it. Great. So build out the first round of this app. And it's actually it was used by both the one here in the Northwest and then also the Colorado Avalanche Center. And so it was kind of this prototype learning. But the main point of it was how do we getting this information back? At the same time, I owned some bakeries here in Seattle and had built out a distribution chain to coffee shops and grocery stores. And similar to kind of the back country problem, I didn't know what was going on with my products in all the grocery stores. And so how can you get that information back? And so kind of combining these things. And talking over a beer, we were talking about how can we get information back and originally we were actually looking for what's known as a it's on Prem, which I know if you encountered that near your thing, but it's it's basically consuming products that consumed on site or on. So like beer from a beer tap and beer companies spend a lot of money doing certain types of advertising. They're not allowed to do some other. They can like having a really nice tap, the the coasters that you're setting your drinks on, maybe the umbrellas that sit outside on the sidewalk, like how how could they get information back on if people are actually doing the advertising they were wanting? So we started down that path to begin with and looking like how can we capture all this type of information using normal people with their cell phones? And what we found was we started getting our traction was from not the not necessarily in these bars and that sort of thing. But we started getting in grocery stores really fast. People did not know what was going on and we knew that market was there, but we thought the other one might be easier to start with. Turns out the grocery stores, the lack of information that these consumer packaged goods brands had was was something that they really wanted. They wanted to fill that void.
Shikher Bhandary [00:10:57] Isn't that weird to think that they made such big moves in the 70s and 80s and stuff with regards to technology, adoption and then. They were advertising all of this, but they didn't know what was coming back, it's kind of like the Facebook ads marketplace where you just put ads on anything and you don't know if it's actually sticking, if it's if it's making any difference. That's why the whole Super Bowl thing is such an important event, because regardless, it makes some headlines, right? You know, there's some impact, I guess,
Erik Chelstad [00:11:33] you know, that's that's like a whole thing, you know, going down that path of of marketing attribution. Right. And I that's something I often liken a lot of what's going on in supermarkets and grocery stores. It's like that old school advertising where they're not necessarily certain what's going on and what the impact is, because it's a more it's a really dynamic industry and dynamic things. How to use your Super Bowl example. It's each one of us had a company and we spent a million bucks for a half second or a half minute or something like that. And so we all go in and, you know, we're all competing against each other. And so not only do you have. Now, this in supermarkets, it's kind of like, OK, so you're putting an ad in the Super Bowl, but actually nobody at your company gets to watch the Super Bowl. So did it happen? You don't know if your ad got shown necessarily
Jed Tabernero [00:12:27] for brands, right. Brands are getting their products across markets and including the online markets and grocery retail brick and mortar stores and whatnot. How did they use to determine the success of a product within a certain grocery chain? You know, beyond looking at, oh, we sold X amount of of products in the grocery chain versus online in the brick and mortar space. Like how did they usually measure the success of us putting it in Safeway or putting it in advance,
Erik Chelstad [00:12:57] you know, matching the post to the point of sale data. That is the main method by which most brands are measuring their success within stores. They buy that data. It's it's actually it's known as syndicated data. And there's providers there's one called spin's and one that we may be familiar with called Nielsen. And Nielsen is somewhat famous for. You know, we talked before about Super Bowl ads. Nielsen is the company that measures how many people were watching that Super Bowl ad and provides that data. So they also purchase the point of sale data from grocery stores, aggregate it and resell it. So it's important to bring that up because this is actually a revenue channel for grocery stores. And so they're able to make a little bit of money selling that data from there, from their cash registers, if you will. And then. So then when they so then brands go buy it and you can now just a couple of points here. One is that's old data. At this point, you're buying data and it's often delivered in chunks like monthly, quarterly, maybe even weekly, if it's something they can do. But you're buying old data and it's kind of like reactive data. It also isn't always it's usually sort of it's aggregated to a higher level. So we're looking at banners or chains, if you will. They're often they're known as banners again on those weird industry terms. But you mentioned Von's. You know, it's like, OK, so I'm going to buy all the sales data for my products at Von's or I'm going to buy all of my category data. So if I'm selling drink mixers or potato chips, I'm going to buy all the potato chips sales data so I can see how my competition is doing as well. And that's that's good data, right? Like you want to be able to know there's a new competitor that's coming in and and, you know, I trying to avoid a bad pun. So I'm not going to say eating up the snack market or chewing up the snack market, but they're they're taking business than that. So this is you know, that's how they traditionally do it. Now, there's other ways you can do it, right? Like looking at repurchasing. How often is is this is a banner or chain buying from you? But remember, now we're talking there's this there's this level of disambiguation. If you go over, it's like I'm getting data, I'm selling to Safeway. And so Safeway's buy my product. It's going to a central warehouse. I don't know which stores are actually getting my stuff exactly if I'm looking at inventory data. So I don't know if it's more popular in Kansas or if it's more popular in Colorado because maybe they're coming out of the same warehouse or just the same chain. So there's there's a few methods like that that people have done. And but that's that's how people used to do it. I got to say, there's there's a great story. One of my favorite ways of doing this was early on, we picked up this customer and they like us. We're a growing new company. And the owner of that company was actually buying bus tickets for his cousins and sending them off to go look at his products and see how they were being represented and count them in the different stores.
Shikher Bhandary [00:16:16] Was this like in the 90s when hundreds. Like how old?
Erik Chelstad [00:16:20] When I was about three years ago.
Shikher Bhandary [00:16:24] OK, so you can definitely see what the issue is. We're very like, you know, you don't get actionable data. About your product in real time, basically,
Erik Chelstad [00:16:37] and you're not getting you're also the data you're getting is it's from it's from a cash register perspective. It's not from the customer's perspective. This is becoming more and more important. But the customer needs an amazing experience.
Shikher Bhandary [00:16:51] Say a brand has a new, you know, packaging. Metha new packaging like, say, Lei's now comes in a box instead of a check back packet right there, don't it's shocking to me that they do not have any insight whether how many people have picked up our product and put it back down and things like that. Right. Because that is important. They picked it up. So that first decision is made. This is interesting. I want to kind of see what this is. But what has resulted in them putting the product back and choosing something else? It's while that that that information's not available to brands even till this day,
Erik Chelstad [00:17:32] you're dealing with a couple of you know, it's expensive to to capture that data because that the way one of the things I like about this problem and it's so exciting trying to solve this and being part of a solution is, is the scale of the problem, because as you mentioned, people are picking things up and putting them down. And what you want is for people to pick things up. Now, of course, you don't want them to put them back down and put them in their cart, but. But they will put them down and they so you got I always think of it like I think it is this funnel where you have, you know, one brand or one manufacturer maybe. And then and then it just starts going down and getting bigger at each level of this inverted funnel, kind of like it starts with this brand and then it goes down to manufacturers. Maybe there's different facilities and there's distributors and and then there's retailers. And then the retailers have like managers within the store and the managers and their stalkers. So at this point, you know, you've got tens of thousands of people, maybe hundreds of thousands of people interacting with your product before it even gets in front of your customer. And then you have, you know, in the US, hundreds of millions of customers going in and hopefully depleting your product, taking it off the shelf or putting it back or putting it back someplace else or knocking off the shelf tag or putting it back on and upside down. Or, you know, however, is being done. There's people interacting with things. And nobody people don't generally I mean, as we're seeing and especially as sort of kind of A.I., some of the the technology we're all part of is developing ways to capture some of that information. As we're seeing, people don't necessarily want that information captured, like they want to be able to pick up a product, look at it and put it back down without somebody knowing that they picked it up and put it back down. And, you know, there's there's been ways that are, you know, people are putting things in or they're measuring your exact location in the stores using different things, whether it's like a like beacons within the stores. So they can see kind of like your dwell time, how many people were there, how many people put in front of this. But there's a lot of information like that that you just mentioned that that is not being collected, that is being collected when we do e-commerce. What a lot of this is, is retail, brick and mortar, specifically using the lessons and the things learned in e-commerce and becoming, you know, as e-commerce has grown and people have done things and learn ways to do it. Brick and mortar is learning from e-commerce coming up with their own methods, but also learning from e-commerce, both because there's great things to learn there, but also because the people that are doing brick and mortar marketing are coming from e-commerce now. And so they can't believe you mentioned it. They can't believe they don't have that data. So who's going to give it? Obviously, this is what I'm doing. I'm going to give them more of this data that they're used to.
Jed Tabernero [00:20:31] That's that's literally like what I was going to ask where all the people that were coming from to solve these problems in the brick and mortar stores, because we know how long something stays on a card online. We know what they looked at, where they scrolled at from e commerce. And like right now, brick and mortar is lagging to get that data and be able to use that for actionable insights. Right. It's fascinating to see that world like coming together. But, you know, as we've mentioned, like, why is this data important for for e-commerce giants? This data is important because they want to improve the customer experience while they're buying the products online. They want to find out what matters, what doesn't matter. Same thing of what you're trying to find out for the brick and mortar stores. Right. So I guess, like, this is a perfect transition. And to me, asking, like, how do you see observe improving the customer experience for a shopper coming in? The brands will optimize and optimized optimize with what they're seeing with the data that they're collecting. But how does that translate? How does that data translate into making that customer experience? Great for me coming into the store, looking at my monster drinks?
Erik Chelstad [00:21:33] Yeah, OK. Well, I see I see two really kind of prime ways that we do this. And I was a customer. We were working with Cona Brewing. They're based out of Hawaii, you know, and they're their companies selling beer. And they were the beer of the month for a chain of stores across California. And they made a big effort with this. They obviously negotiated with this store chain and to get this beer of the month thing, which meant specials and also met these really beautiful displays that their marketing team had built out. And these were they have their spokespeople at the time or these two really big guys like NFL linebacker size guys with. So they made cardboard cutouts of these guys and they had full like real world surfboards up there. So this huge display stacked with bears and informational display about this. And so what this was doing, I mean, this is part of the customer experience, right? This is one of the reasons you like to go to stores or we do as people is that you get to see these things. And it's a cool it's a beautiful display and it's interesting. And maybe you're going to try this beer. Maybe you're already a fan and you get it and you're like, oh, I love this. And there's a great thing and you have a good experience because you got to interact with your favorite brand and see this beautiful display. Or maybe you're just getting something new and you're going to your friend's barbecue and you talk about this. And so these are the. Types of things that, you know, we go to. This is one of the reasons we go to stores. So what we did is we went out and we found out that in the first to one month before the month thing, the first weekend, only 20 percent of the stores had actually set up these big, beautiful displays. And so that no one gives our customers something to work with and gives them not just a generalized 20 percent, but they knew exactly the 80 percent of the stores they needed to contact. And we're working with our distributors, et cetera, to get these things built out and set up. And now that that also led the marketing team to understand potentially why their giant investment in all of these displays did not have a marketing need to not have a sales impact that they wanted it to. Because you can imagine the disappointment when you build something and it's amazing. And then you send it out there. And two months later, when you get your your data back, you find out it didn't have it didn't work. And you don't know why. I liken it to somebody doing a B testing on a website and not knowing if they're seeing A or B or even see some other variation of the website. It's just so that that marketing provides good customer experience, which is just meant to do, then that is one of the customer experience, things that we help out with. And then the other the other part of it is just making sure that the things are there, that the products are there. And that case, what we do a lot of is just validating that the products are actually out there on the shelf in the way they are supposed to be. So it's not always the merchandizing, it's not always even planogram compliance. And a planogram is the design of the shelf down to which products go work on each individual shelf, which is also something that people put brands put millions of dollars into researching and deploying. But so these are all the things that you want as a customer. I want a good experience. I want to know that the product is going to be there when I go shopping.
Shikher Bhandary [00:25:03] Shelf space is so important, right? Because the most important goods are at eye level in a grocery store. Right. That's common. Like when you walk in a grocery store, you will see crafts catch up at your eye level and down below will be like a boutique brand, which is like 100 times better than craft schedule. But that's how the system works. Right. So the grocery stores, you know, making it even harder for brands to actually actualize that data by having their own label brands, having their own agendas, having their own revenue streams coming from these white labels. I guess
Erik Chelstad [00:25:39] so. The brands that is a competition. There will always be competition. And there things I think it's interesting, you brought up the eye level and you mentioned, like, if you're looking at ketchup, that it's Kraft, Heinz in that right there at eye level. Kraft, Heinz is paying for that. Yes. They're known as category captains generally because one that they're the biggest. They're going to be the biggest brands in each category. The store, they they create, you know, agreements with these other companies. And they're often the ones they're going to be, the ones that design that entire ketchup section for all of the brands. And so they're spending they spend tens of millions of dollars to design this and they're using augmented reality testing. Kind of go back to your question about picking things up and putting them down. They're doing this and they're creating these these optimized planogram that then roll out to these stores. And and so the brands are you know, that this is part of what we do is we ensure that those brands are getting that coverage that they that they've agreed to, OK? And so do they have the right shelf space. And, you know, they have these metrics like share of shelf, which is, you know, how much when I look at the shelf, how many linear inches does my product have versus my competitors or versus everyone else. And you can kind of look at share of assortment, which is another one, which is how many products do I have versus my competitors versus space. And so there's always these games going on, you know, where people do you want to take up more space or less space and you might want to take up more space if you are a category captain and people come into the store to buy your specific ketchup or your brand of hot sauce. So you won't want to take up as much shelf space as possible. But if you're an upcoming brand, you might want to take up the least amount of space as possible. So they will slot you in there and you can get three different flavors in the space of somebody else's one flavor. So it does there's a lot of interesting games and the kind of go on and in that serve the real estate of the shelves. And then when you start talking about frozen goods, it's even it's a crazier game because that space is in more cutthroat. It's more expensive to me
Adrian Grobelny [00:27:52] to be in the frozen game. Yeah. Through doing this research, I was reminded of a case study story that was really funny and interesting. And one of my business classes, and it was that and in 1992, this consulting agency, I don't know which one it was, did a study for grocery stores to kind of find commonalities and trends that are happening with grocers and why and they're trying to understand the psychology behind it. Why are certain items bundled together and why are certain items never bought together? And one interesting correlation that they found was diapers and beers. There is a point nine five percent correlation of those two items being purchased between I think it was like eight p.m. and 12:00 p.m. or midnight. And they're, you know, they're scratching their heads who's buying diapers and beers at night with a point nine five percent correlation. And it end up being, you know, the wives asking the husbands, get get some diapers, get we need more diapers for the baby. And they're like, all right, I'll get the diapers. And then they think I need to get something for myself. Well, I'll get some beers while I'm at it. And so that was a really interesting insight. And I haven't really seen lately any stores bundling diapers and and beers together like they do with s'mores, you know, the marshmallows, graham crackers and chocolate together, which is really common. But that was just a really interesting insight that I think no one would have ever thought could have correlation. So what what are kind of the types of insights that you focus on gathering, you know, without spilling the secret sauce that you guys are working on and and, you know, staying competitive? What kinds of insights are you providing businesses and how do you really use eye to package it and make it understandable for, you know, a brand that is not technology driven like your companies?
Erik Chelstad [00:29:44] Have you noticed what's going to ask you? Of course. Have you seen what's being bundled with beer when you go to the grocery store lately? Chips, if you look around, you're in the beer. Look look around what's there. But you'll probably see beef jerky in that section. But the one that I really love that's showing up a lot over the last couple years is ping pong balls.
Adrian Grobelny [00:30:03] Oh, no, that's so true
Jed Tabernero [00:30:05] because of Adrienn. That's because of you
Shikher Bhandary [00:30:08] doing that is how
Jed Tabernero [00:30:09] many frat parties online
Shikher Bhandary [00:30:13] their number one customer for that bundle.
Erik Chelstad [00:30:17] So so it is interesting in those types of insights. Those are great. And, you know, we're get so we're providing something that's a unique data. So people are coming up with their insights on what what that might be. But, you know, a lot of it is. So some of the some of the more basic stuff is is like and I say basic because it feels basic, but it's really not because the information wasn't there before, but kind of trend analysis. Looking at when we talk, when I was mentioning before, like a share of shelf, when you're measuring how many of your products are on the shelf versus your competitors or how much space your products take up versus your competitors. And if you watch that over the course of 12 months, if you've been measuring it every month, every week, and you're getting that insight and you can see the trend, are you getting more or less? And how does that correlate to your sales? It's also really interesting to note, you know, when we talk about so some two of the big problems that everyone in CPG faces are out of stock, which is when obviously the one seems pretty self-evident, they're what that is. But there's another problem called the void, which is similar to that of stock, except that basically the product that you're allowed to sell at that store or that store is authorized to sell is not on the shelf. So if I have chips and I have, you know, like if I have salt and vinegar and how a penny of flavored, then if, you know, the stores don't have housepainter, they don't even have a price tag for it. So it's not a stock, it's just not there. These are two pretty major problems. And so. You know, those those insights like capturing that over time is interesting because it might mean if you're always out of stock of something, it might mean that your product isn't getting to that store at all. But it also might mean that it's so insanely popular. The people are just buying it and buying it just flew off the shelf. Now, you might not be able to see that in your order in that sales data because you're just going to see that you sold out. But what you don't know is that you sold out by 8:00 a.m. every morning. And so that can be used. And so this is something I think is important, is that that data the retailers want that data as well. And when the brands bring that to them and say, hey, my chocolate ice cream is out of stock all the time and we want to get more of it on the shelf so it doesn't run out, the retailers are receptive to that because they know that they're going to make more money because that volume is increasing and so everybody wins in that case. And so that's the kind of thing those those are the types of insights that you can get pretty easily from like by doing what we're doing and analyzing every product
Jed Tabernero [00:32:58] I wanted to to go directly to the observers. Like if you could help us kind of understand what that economy is like, like who drives the data collection, whether it fits like something that somebody sees a page and says, hey, listen, this brand wants to check out there. They're shelving for the specific product at the specific store. Is it driven by that or is it driven by, hey, I'm going to the grocery store like every week and I've made a commitment and I'm going to take pictures of a bunch of brands that that are part of this.
Erik Chelstad [00:33:26] You know, our customers of the brands primarily, we do work with retailers as well, but it's a brand and we'll make it a fictitious brand that's checking on any flavor of mayonnaise or the mayonnaise section, which, by the way, is way bigger than you would expect if you. That's another thing. When you go to the store, try to find a section where there aren't like 100 options for any product. It's amazing. We've done pretty well with expanding on products, but so our customer is fictitious mayonnaise company and they want to know what's going on on the shelves and the stores. They might have a goal in mind. They might be trying to renegotiate with the retailer buyers to get more shelf space for their new flavors or a bigger jar or whatever. And so then what we're going to do is we're going to say we're going to basically create these opportunities is what we call them. And those are specifically all these locations across the US. And so we put those into our system and they're going to launch on a certain date because we don't want to go before or after. You know, there's usually a time window that you're trying to capture this data. And so then that when it goes out, it goes to basically this all of our observers that are anywhere near that are going to get notified of this. And so we've got about three hundred thousand observers right now in the US. And so they they're going to be notified of that if it's near them. Wow.
Jed Tabernero [00:34:57] Yeah, that's
Shikher Bhandary [00:34:58] you. I thought you were expecting to say three thousand and you're like hundreds and said three hundred thousand of those like we had. Did he say three hundred thousand. So three hundred thousand people. Observers I guess are the folks that go into the grocery stores and collect the data, basically take the pictures that your software. You know, yeah,
Erik Chelstad [00:35:19] yeah, OK. Correct. So they're there, they're out there and there's two primary groups of people that that work with us to do these observations, it seems to be there's people that walk into the grocery store and now the tag actually, you know, we made it. So it's like go away. So when you're actually right next to a grocery store, you'll get a pop up notification, says, hey, you're within 100 feet of the store. Do you want to make five bucks? And so this is our primary motivation is we're paying people, paying people. We're paying them as fast as possible. And because that's a motivating thing, this is you know, it's a big deal. It's hard to create, you know, creating a marketplace like this. And, you know, at the very beginning when we were talking to potential investors, it was always like, well, this is a chicken and the egg problem. How are you going to solve this? And so for us, it was treating treating everyone well. I mean, this is kind of basic customer service, but it's not always something we see, you know, in the industry. When you're creating a crowd, treat them the best you can. Be transparent, be quick about paying. But then so, yes, they go out, they collect the information, they're taking pictures, they're entering in numerical information. So we capture your photos primarily. And then these photos come back to our system and we recognize them using neural networks and image object detection. You know, we're targeting and we're capturing, drawing, bounding boxes around every single product on a shelf. And so we're going down to a package level. So we've trained our models on the actual packages. You know, we have hundreds of thousands of observers. So when we need training data for our models, they can go get it for us. And and that's the generally that is what you want for object detection, as you want as much data, as many images as possible so we can get that data, train up our models, keep them fresh as packaging changes. Because you mentioned before, Lei's going from a bag to a box. That would be a pretty massive change, but there's a ton of little changes that happen.
Adrian Grobelny [00:37:23] It's really interesting that you still rely on human observers to get all this data. You know, it's actually timely. Wal-Mart actually didn't lay off 500 workers. They laid off five hundred of their robots that they were using to drive through aisles, scan the aisles and then basically collect data continuously. It was it was like a contract with bossa nova technology. And they basically, you know, through this three year contract that they had with this company, they found that they're getting the same results, that they were with humans, with these robots. And so it's like it makes you scratch your head and think is is the is that technology ahead of itself? Is is there you know what? Why why is this new technology kind of, you know, taking a step back? And why are we still depending on humans to collect this data?
Erik Chelstad [00:38:17] It's a great question and it leads me down this path of so we have the observers. This was when we started the company. We knew that we wanted to get this data and the majority. So retail, you know, it's a thin margin gap. It's a it's big business, but it's thin margins. And so, you know, we're talking single digit percentage margins for retail stores. And I mean, a lot of, like, stores might be operating on a two percent margin. You know, there's these things. It's very slim margin. So they're always seeking out ways to do things better and a little bit cheaper. And obviously, you know, real estate and labor are you're going to be a big cost. But we don't we wanted to get people out there. We knew that, like going out, trying to get retailers to move on. Like our technology solution might take a little it's going to take a little bit longer. Brands are the one that has a lot at stake. They're able to move faster. They are innovative by nature because they're constantly competing with each other. And so so we would go after them. But we got people in there because we want to use people and make it happen. We have partnered with robot companies, automated drone companies, shelf camera companies are that's that was something we we did not want to get into that game. I don't want to make those robots. I mean, personally, yes, I want to make those robots. I'm an engineer. I do want to do that. But it's not that wasn't the business we want to get into. We we partner with those types of companies like bossanova in order to provide something else that they're not doing. Even, you know, some of our customers actually use their own employees with our system because it allows them to optimize their time in the store. If somebody is going into a store and you have two hundred products on the shelf, you know, think of some of those. Think of makeup and vitamins and these types of things where there's just a bunch of different options from one manufacturer. Somebody can go into the store. Take a picture, get the results back from us and know what they need to do, which means that their visit, if they have 20 minutes allocated for a store, they can spend five minutes doing, you know, the kind of merchandizing activity and then they might be able to spend 15 minutes talking to the manager and selling new products.
Shikher Bhandary [00:40:38] We didn't dove completely into it. But if you can just outline how that how Observer's technology works from start to finish. So the different stakeholders. So now you have an observer, they come in and take a picture that gets fed into a database, which you guys obviously run your algorithms and you get actionable insights from that, which then gets relayed to the brands that, you know, can observe trends and stuff. So is that the whole, you know, cycle of your product?
Erik Chelstad [00:41:12] So the basic. Yeah. So I mean, the kind of that that life cycle of, you know, the basic life cell and this is the one I like that to focus people on our customers is that we create a campaign and we try to use words that are similar to, you know, familiar to people in these industries. So it's a campaign. We collect the information, we do the observations, and as each one as it comes back, becomes reportable. And then there's also a reporting later, aggregate reporting on the on the entire campaign. Something that we haven't really touched on is that we also when you asked about people, one of the nice things about having people is we have them out there and we can actually do more than just do an audit of what's happening on the shelf. We can actually fix problems as well. We can have people do actions. So those actions might be I mean, it might be as simple. They might be rearranging the products on the shelf. They might be actually talking to a manager about reordering products and those types of things as well. So having people out there gives us that flexibility to do that. It also allows us to pay people more. And that's always something that I like to do. It's one of the things I'm proud of is that you get we get people that talk to us and observers that that tell us how, you know, we've made an impact in their life and some positive impact, I should say. And it's and that's really nice to hear. And it's good to know that that's going on. One of the things that doesn't always come up in like a sales pitch like that or, you know, going over the cycle is that that we're validating the data. And so it's an important piece of our technology stack in that when the data comes in, we're doing a lot of basically fraud detection in the same way that banks are doing it with your credit card transactions. And we're looking at a lot of all the information that we can make sure that that we can get to make sure that somebody is where they are supposed to be and is actually doing collecting information correctly. Because a lot of times, remember, we're dealing with problems like the question might be, are my products on the shelf at this location in Schenectady, New York? And if the answer is no, your products are not here. We need to be able to make sure that that's correct. And so we need to make sure that the observer was there and that that they were actually in the right place at the right time, looking in the right area.
Jed Tabernero [00:43:45] And yeah. And, you know, just to touch on, like, the ability to be able to scale technologies like this, people don't really understand yet because there's there's no canonical stack for four Mellops, you know, like it's a whole we just did an episode on data labeling company. Right. And we just oh my gosh. The stuff that we had to learn about Mellops and how it's getting developed right now, it's ridiculous, especially for somebody who's not in the tech space. Right. So there's still tools that need to be built to be able to handle this kind of scale, especially what you're doing with, like the iterations, getting customer feedback, getting feedback from the observers and whatnot, like there's a lot to be built there and supervises an exciting space to get to next. I've learned so much with this episode, just about the brick and mortar space and about how artificial intelligence and machine learning is interacting with us right now and like how this new goldmine of data is being collected from the actual point of sale. I would say like the actual place where people are looking at products and deciding whether they want it or not. This, you know, the point of decision. There we go. So so not necessarily a point of sale. So there's the pure system data and there's going to be a point of decision data which is coming from service. So that's really exciting. And then also the fact that there are observers who can come out and do campaigns for specific product launches, for specific data points that they want to gather to be able to create actionable insights. You know, so it's fascinating to see and I've learned so much throughout this call, but, you know, I want to be able to to point our listeners to a specific place or. Somewhere where they can reach and learn more about you and learn more about observe. And, you know, I think a lot of the listeners that we have today are increasingly getting exposed to artificial intelligence conversations. And so we always point them out to resources on where to learn about this kind of stuff. And we're learning about that and observe his website has some resources to be able to understand it. So, yeah, if you could point our listeners to somewhere that they could go and learn about your product and about yourself, they'll be great.
Erik Chelstad [00:45:47] Yeah, I mean, I think, like, our blog post is a great place to go on. Observe now, dotcom, the the blogs are under the resources. They're digestible snack sized blog posts about AI and retail. And we pay our observers and through PayPal, but we also use crypto currencies if they want. And like why we did that. And I try to I try to introduce people knowing that there are a lot of people in you know, retail is not a technology game. It's becoming one. But it's it always it hasn't always been one. And so I try to make things, you know, speaking to intelligent people that aren't exposed to this every day. I do like talking about this stuff and not everybody wants to hear about it. I would say that the largest change that we're having is to allow our customers to make informed decisions based on real time data. That was something they could not get before. And I say this because obviously there this can expand beyond retail. You know, it's creating a distributed system for gathering information that is reliable and actionable. We are making it so retailers and consumers, consumer packaged goods, brands, are able to make real time decisions and and provide a really solid customer experience.
Shikher Bhandary [00:47:08] Hey, thanks so much for listening to our show this week. You could subscribe to us and if you're feeling generous, well, you could even leave us a review. Trust me, it goes a long, long way. You could also follow THC @thc_pod on Twitter and LinkedIn. This is things have changed.