Shelf Engine manages the perishable food orders for the likes of Walmart and Kroger by streamlining the ordering process so the right amount of product is on the shelf at the right time.
Seattle-bases Shelf Engine, which raised $41m (£29m) earlier this year, valuing the firm at $110m (£78m), is looking to wring inefficiencies out of the supermarket stock ordering system.
It does this by providing what it calls "demand forecasting" for perishable food items. Shelf Engine cuts down on supermarket wastage by using its AI technology to ensure the right amount of stock is purchased to meet customer demand, and not too much. Shelf Engine guarantees set sales for a supermarket, and perishables that are unsold it pays for.
In this interview, Stefan Kalb, co-founder and CEO, talks Future Food Finance through the business model, telling us that people thought he was "a bit crazy" to set the business up but that through its technology it now "knows what's happening at the grocery store better than the actual grocery store".
Kalb also tells us about a famous early investor in the business, its growth plans, and how it will look to move beyond perishable foods.
Paul Cuatrecasas: So, Stefan, kicking off with the basics here, what is the 30 second elevator pitch of Shelf Engine today?
Stefan Kalb: Shelf Engine does something rather unique. From the outside perspective, it looks like all we do is demand forecasting for highly perishable products. But, what we actually do is we guarantee sales for grocers. By doing the demand forecasting, we take over categories at the grocery store and we literally manage the ordering, and we guarantee the sales for the grocery store.
PC: So, you said you “take over” orders, is that right? How do you do that?
SK: Essentially what happens is we come in, and we go into a given category, and we say “we will order all the products from that category”, we will pay the vendor for everything they deliver, and we will only charge a store for what sells. In addition to that service, we also go to the grocers and we say “we will charge you a new price for the same item that you're currently carrying, and that new price that we're going to charge you is going to be way less than your current shrink rate. Which means overnight, you're going to make a meaningful increase in your gross profit”, and that that is our direct sales pitch.
PC: So, your pricing, your margin comes from capturing the shrinkage, the spoilage, the waste that they're currently experiencing, so that they effectively have a higher margin, even though they might be paying more on a headline basis, they have a higher margin and you are also capturing part of that margin as part of the service provided?
SK: Yep, that's exactly right.
PC: Fascinating. Are you aware of other companies that are doing this anywhere else in the world?
SK: I'm not. What’s really interesting about this is that when we first came out with this model about four years ago, at first people thought we were just a bit crazy. But now what's interesting is we're being called this kind of new concept, "outcomes-as-a-service". And I never knew, especially when fundraising, what to really call what we did, right? Like, we're not SaaS, we're not a marketplace. We've got some aspects of these things. Outcomes-as -a-service is maybe the best definition I've heard thus far.
PC: Where did that term come from?
SK: I'm not sure where it originated. I think there was some research article written at one point about it and then it finally became a conversation point.
PC: I like it. You mentioned how it's unique in terms of guaranteeing the sales for the grocers by taking over the orders. How are you able to outperform the store so much? Why are you able to do that?
SK: There’s two reasons for that. The first reason is that we have access to data that nobody else has access to, that gives us a huge leg up, and that data is collected by our own team. We have a national field team that goes into stores, goes into vendors, and collects data.
And effectively, we know what's happening at the grocery store better than the actual grocery store knows what's happening.
The second thing is, in terms of our forecasting abilities, frankly speaking, we just have the best forecasting abilities I've ever seen. We got started early enough, and we've been growing what we do at a high level. It's obviously gotten to be very complex, but it takes in these multiple streams of data and it's able to essentially forecast every single SKU of every single store every single day.
It is outperforming any human and any other system that is out there, and so that's really what gives us the ability to price so aggressively. We go to a grocery store, and we say, “hey, in any given category you have, we will increase your gross margins to such an effect that there will be more than anything you can ever imagine".
PC: Any category?
SK: Yes, the categories that are going to have the largest impact are going to be the freshest ones, the ones with the highest spoilage rates. So, take, for example, the first three categories you'll see when you walk into a store, which is produce, deli and bakery.
On those three categories, on average, we are increasing the gross profit for the grocers by an additional 10 points. That means it's not unusual for gross profits in those categories to be, say, 20 per cent. And with us, now they're 30 per cent. So, they're absolutely massive increases, right? And the way that we're really able to do that is by outperforming their forecasting abilities.
PC: So, you mentioned the access to data and how it’s collected by your own team. I recall a discussion I had with a very senior executive at a large supermarket operation in Europe a few years ago who said, "we have so much data, but we don't know what to do with it". And I've also heard this from other companies that “we have huge amounts of data, but it's not very high-quality data". If it's not high-quality data, there's not much you can do with it. So, do you make a major effort to ensure the data you have is of the highest quality?
SK: Yes. There's essentially two ways to think about it in terms of the quality of the data. It's absolutely true that most grocers have mounds of data, and they don't know what to do with it. In fact, it's very siloed in different parts of the organisation. One of the key things we noticed is that most grocers actually don't know what the true shrink rate is. That's simply because their sales data literally lives in a different system than their delivery data, and it's very hard for grocers to come up with that information.
So the first thing we do in any of our processes in terms of the sales process is we say to the grocer, "for free, we will do analysis on your data and we will tell you what the shrink rate is. You don't have to work with us or do anything. We'll just tell you what the real shrink rate is". And I actually have had several fairly uncomfortable conversations where very important executives thought that their shrink rate was much lower than it actually is.
PC: Why would anyone say no to that? It seems like an irresistible offer...?
SK: Yes, well, we get kind of yeses on every level there. And so what's happening in terms of the data sources and why there's so many issues is really because there's one source of truth, which is the sales data. And the sales data even can be corrupted, but generally speaking, that's the one source of truth.
Almost everything else in the store is going to be managed by humans, which means that is when you're going to start getting a lot of issues. The opportunity is to use all that data to get to a source of truth.
So, for example, one of the major advantages that Shelf Engine has is that we can predict the inventory in the store way better than anybody else can. This is part of the reason why we're so strong at ordering, not just because of our demand forecasting, but also because we can predict inventory.
We use the grocer’s data sometimes to do this, but most often it’s from our own field team, and that part takes some very complex AI to be able to forecast accurately. A grocer could, in essence, put together something to get that same data, but it would be a huge challenge on their part.
PC: Interesting. Why isn't every grocer using Shelf Engine today?
SK: They will be. We are early in our phases. We're growing as fast as we possibly can. We’re currently working with some of the most notable brands out there. We’re onboarding new stores and new categories as quickly as we possibly can. The thing is, there is no possible future for grocers without Shelf Engine or a Shelf Engine-type of company that manages all the ordering, it just doesn't make sense to not do this.
PC: And what is the percentage of sales today of US versus Canada versus Europe?
SK:100 per cent US. We will have our first launch in Canada shortly. We are going to be sticking to North America for a while. We'd be very excited about Europe but it's quite an investment.
PC: So, you've got the capital, what is the greatest barrier holding you back? Is it sales and marketing? Is it selling? Is it creating demand? Is it literally just blocking and tackling right now to grow?
SK: So, there's essentially one major aspect that we're investing in, which is how to scale. What we do is really complex. You can imagine that basically, every forecast and everything we do lights up an insane number of servers to be able to do that. So, how do we create an extremely scalable solution to be able to grow from 2,000 stores to 20,000, 40,000, etc? So that's a big focus for us right now. We’re investing heavily in our data science and the engineering teams to be able to pull that off.
PC: Let's rewind a little bit. What was the original inspiration for you to set up a company with your co-founder? Was it the follow on from something else…?
SK: Yes that's exactly right. My first company was a food production and distribution company. We worked with a lot of foodservice companies and a lot of grocers and I learned how hard it is to actually forecast how much food to order. When I went and learned the real percentages, what the true shrink percentages were, I was pretty shocked, and that was the catalyst.
My co-founder was at Microsoft for about a decade and I have a background in math, so when we were chatting about this problem one day I said “I could write the math portion, I could write a model to start forecasting this, and you could create an app and let's go and test this and go about it”.
And interestingly what happened was we started to go to coffee shops in the evenings and just kind of worked on it together and it was a good time, then we got really fortunate because we were actually still both at our own jobs and we got some financing options.
There's actually a pretty good story where (American football legend) Joe Montana was one of our first investors, and so we had a couple of those situations where Joe along with some other folks were like "well this sounds really interesting I'll give you some money and you go and do it". So, we were really excited for that opportunity and that's kind of how things kicked off.
PC: Who were the first employees that you hired and how did you get those first people on board?
SK: Yeah that was hard, that was really hard. So, one of the first was an engineer that (my co-founder partner) Bede had worked with at Microsoft and then a data scientist, and then I was out there just selling nonstop. It was a bit of a slow start because the sales cycle is so hard, it takes a while to get to any sort of scale.
PC: What was the first major sale that had an impact, if you can talk about it?
SK: That would have probably been Whole Foods. That was quite meaningful for us in terms of a sale. They're an interesting opportunity because unlike many other grocery chains, they're very decentralised, which gave us an opportunity to sell in and start scaling with them.
PC: Presumably you had to say "we'll prove it to you"?
SK: To be honest we still do that. I think the grocery industry has a lot of sceptics and so the number one thing we do when we go into a new store is we say "give us five stores we will transform your bottom line. You will be able to see the impact across your entire chain'.
PC: Interesting. Where do you see Shelf Engine five years from now? 10 years from now? Knowing that there's going to be huge amounts of change in the overall food value chain? I'd be very interested to hear your view from a retail perspective, and also from a supplier’s perspective, where you see the company in 5-10 years’ time?
SK: I'm glad you're asking because most of the time people are asking about everything that is very consumer-based. They're often asking about Instacart and what's online grocery going to do, and I actually think the thesis here is that there is going to be a much stronger development on everything that is the back end rather than the front end of what's happening in grocery.
From our perspective what we're focusing on is saying "how do we manage as many of the categories and many of the stores as possible?" Because then we want to start offering things up the supply chain that are going to be very beneficial. Now we're basically in charge of the waste, we're heavily incentivized to make sure that we're reducing it as much as possible, and increasing sales as much as possible, so what can we do with all the distributors and all the vendors to be able to make everything really efficient.
We're already working on some things where we can guarantee production amounts way further in advance, so that vendors can prepare, have more product on hand and potentially give better price breaks. Then there's a lot of other things that we're looking at on that side of things, saying "well if we were working significantly through the supply chain what can we do to add significant benefit to our customer?"
PC: Now let me ask something a little more radical. If we go out five to 10 years, and we don't know the exact timeframe, but a couple of forces that are emerging. One of them is D2C from new brands, so for example, Beyond Meat and Impossible Foods, building a brand and going direct to the consumer without necessarily having to go to a supermarket, we're going to see more of that.
But possibly more significantly is the form of alternative grocery and delivery, not just that the delivery providers today are raising huge amounts of capital getting supercharged off the pandemic, but what we certainly see is what nobody else is really talking about yet and that's drones.
I've been talking about this for a few years and I talked with Bobby Healy, the founder and CEO of Manna Aero, which is actually operating drone deliveries in Ireland at the moment…they have a town of 10,000 people, and they have a drone service that's delivering fresh coffee in four minutes. They’re delivering Just Eat Takeaway food, Tesco groceries, and people apparently love it.
Bobby says people aren’t going back to doing what they did “pre-drone". He said that this is not that far off, the Irish Aviation Authority approved service recently and they're close to the European authorities, and they may be approving (drone delivery) in the next 12 to 18 months. He said, in five years, to give you an example of how we will be operating, we'll have a city of 20 million people, and we will be doing half of all the last mile delivery in that city. So only half of it will be by road.
So, I'm just thinking ahead for Shelf Engine today. If today, it's grocery stores and its supermarket offline presence, bricks and mortar, you've got so much to grow, you could just say “it doesn't really matter because we're going to grow into whatever's left”. But certainly, you're going to shape and mould yourself even more to how things are going? Do you have any views on that, longer term?
SK: Absolutely. I mean, the core problem is always the same. Even if there isn’t brick and mortar by that point and it is all drones, and it is all distribution centers then they still have to figure out how much product they're going to put into the distribution center. What do you have? You have richer data, because you actually have the customer’s direct information that you can work with. But the core problem is still the same, and interestingly, when we come into that kind of situation, we're not just a service, we're completely eliminating the risk.
And so, if you think about yourself being a warehouse that's going to distribute to all these places, wouldn't you want the very best solution in terms of making sure that you have the right amount of products in your warehouse and eliminate all of your inventory risk? It's just going to make sense in the long term, no matter what grocery evolves into.
PC: Let’s take vertical farming, which was happening well before we heard news about SPACs and the like. If we take the movement of fresh foods and pure foods, where we have these movies like “Secret Ingredients” and “Kiss the Ground”, and now “Seaspiracy”…it's all building that premise that people increasingly want “pure”, more pure than organic food, and so vertical farming promises that for fruit and vegetables there are certain vertical farm operations that in some cases could be at the supermarket, but in other cases might just be directly sent to the consumer. Is there a model there for you? Let's just say that really does take off in two or three or five years….can you adapt to that model?
SK: Yes, absolutely. I would say the same thing (as drone delivery), right? Essentially, how can we go to the vertical farms and say, we will guarantee all that you produce, and we will essentially know what the consumer is going to buy and make sure that we're telling you exactly what needs to be produced and how much we're willing to guarantee.
So, I think the evolution of a lot of these D2C concepts are super interesting, and we will likely play a part in accelerating them by the means of essentially guaranteeing them and helping them predict how much to produce.
PC: Now that brings up another idea. I know you have to walk before you run and you're probably still feeling like you're in the crawl stage but from everything I'm hearing and everything I've learned and read, it seems like Shelf Engine, especially if you're able to master this guarantee model, can evolve into a company that can actually provide this service for more perishable foods and grocery but for other goods as well. Is that on the horizon?
SK: Yes, the one that we've gotten probably a lot of outreach on is on medical inventory. I’m not a healthcare expert, so I didn't realise how much product in hospitals has a short shelf life. There are things like skin grafts to blood to medicine. And so, we have not mentioned that space at all, but if I was going to hypothesize in terms of kind of a next phase, that'll probably be in the healthcare space.
PC: I was thinking about cultured fish and cultured meat, which is something we follow quite closely and is also gathering quite a lot of steam in the private capital markets as well. Many of those companies are B2B. They are going to build bioreactors, and they're going to work with the retailers. But I guess for you, it's the same answer, isn't it, as drones or vertical farms? It's wherever the goods are that are perishable, like all of that (cultured meat) would be, you'll have more data presumably than ever before, to work with?
SK: Yes, I mean, the variables are the same, which is essentially how much demand is there? How much product is there on the shelf? What is the shelf life of the product? And what is the gross margin of the product? All those factors are the ones that essentially create the profit maximization function for how much product to stock.
And you can essentially say, it could be a magazine on a shelf, it could be, some milk, it could be some sort of cultured fish, but the idea is all the same. Those four components are essentially going to make it some sort of profit maximization function.
PC: Just out of interest today do you have enough data to be able to know what per cent of all your customers in stores that have perishable goods have fresh fish? Do you know what percentage of that you guarantee out of all products?
SK: With a lot of our major relationships we have full POS integrations, which means that we get all of their data, even if we're not managing their entire store. So, we have incredible insights that even if you went to Nielsen, or IRI, you couldn't get the richness of the data that we have. Now, we are very aware of how important the trust is that we have with our customers.
So, we don't share that information, we don't share those global insights, but it is something that we can see on our side, and that brings tremendous value to what we do.
PC: On that note do you ever see the need to adopt a blockchain type of protocol for enhanced security to ensure that your data is secure, as well as your customers' data?
SK: Potentially. We do some really complex stuff on the security side of things. And obviously, my co-founder coming from Microsoft is very experienced. We're built on the Microsoft stack with some pretty amazing security protocols. We may one day go to blockchain, I can't really speak to it completely, but it might be the right solution one day.
PC: Yes, this is so interesting. Stefan, is there anything that you wished I had asked you but I haven’t?
SK: The question that I think about a lot is what happens when grocers start making a lot of money. And I'm being completely serious, what happens in a highly commoditized industry, where we can make grocers a lot more money, right? That's the exciting question. What do they do with that money? Do they reinvest it to lower their prices? And that's the one that would be fun. I'd love to be on a panel with some grocers and talk about that.
PC: Okay, well maybe we can help make that happen! Thanks for your time today, Stefan, exciting times ahead.