Halla Offers Personalized Grocery Curation and Prediction

Technology, technology technology. It's all that the grocery industry is thinking about. How to increase efficiencies better than the shopper experience and create bigger basket size today. My guest is Spencer Price co-founder and CEO of Halla. Halla is the taste intelligence company and creator of the only human preference engine that's designed for grocery. Its tastes intelligence uses AI to redefine personalization and to predict what individual shoppers will actually want to buy next in real time.

Transcript

Phil: 0:15

Welcome to Lost in the Supermarket. I'm your host, Phil Lempert. Technology, technology technology. It's all that the grocery industry is thinking about. How to increase efficiencies better than the shopper experience and create bigger basket size today. My guest is Spencer Pprice co-founder and CEO of Halla. Halla is the taste intelligence company and creator of the only human preference engine that's designed for grocery. Its tastes intelligence uses AI to redefine personalization and to predict what individual shoppers will actually want to buy next in real time. Spencer, welcome to Lost in the Supermarket.

Spencer: 0:56

Thank you, Phil. Super excited to be here.

Phil: 0:59

So you have a very different approach to personalization. Um, what you try to do is understand how each individual customer enjoys grocery products. Tell me how you do that.

Spencer: 1:11

So the best way of describing what Halla does and specifically our tastes intelligence engine, which you described perfectly by the way, that was awesome is really by looking at other platforms out there in the content realm, not necessarily shopping, but let's think of Netflix or Spotify. And the reason for that is because they've got, let's take Spotify specifically tens of millions of unique songs and across those songs that are going to be different categories or different pieces of metadata associated with each track. And so what Spotify does to power highly personalized recommendations, which generate over two thirds of all the music streamed on their platform, by the way, is they break down every single song into hundreds of potential. Sub-components from objective analysis and metrics like instrumentation and arrangement and tempo all the way through to more subjective things like dance ability and Halla does the same exact thing with food. And so we'll take every single product in a grocery stores inventory that we're working with and break it down into preparation style seasonality, occasionally you name it so that we're not just understanding how people relate to products, but what the products actually are and how they're used. And that is Khalid's bread and butter.

Phil: 2:30

So let's take a box of Cheerios. So when, when I look at a box of Cheerios, um, you're, you're pulling hundreds or thousands of attributes of that product if I understand properly. So you're looking at nutrition, you're looking at adding milk, you're looking at putting it in a bowl. You're, you're doing all those things and then taking that information and doing what with it

Spencer: 2:58

Precisely we're taking that, that information and a whole bunch more, including your household habits, dietary restrictions, your taste preferences, cravings, you're likely to have what you might be running out of. And of course, looking at the distinct curios, you've selected relative to all the other options that are out there and what we're using that information for us to understand what the best complimentary products might be to suggest to go with that Cheerios. And that's based on what's already in your cart. So if you've already added milk, we're probably not going to surface in your face a whole bunch of other milks.

Phil: 3:31

Instead, we might be showing you things like banana that you might want to add to your cereal, but if we can glean from your browsing behavior, your past purchase history, all without any personally identifiable information, what it is you might be likely as to enjoy in addition to what's in your cart and the product you're looking at in this case, Cheerios, it might be banana, it might be cinnamon. Um, it could be a whole variety of things. And our goal is to understand that for each unique shopper and present you with the best possible options to make it as easy as possible to add those products to. Um, so, so I see the benefit for the retailer and for the shopper, the shopper discovers other products that they might not be thinking about. Uh, the retailer gets to sell more stuff. Um, if you would, um, what are, what are some of the retailer observations that you've got as you're exposing this to retailers? What are the kinds of questions that they have? And if you could share any of the case studies without naming the retailer of what the results have been.

Spencer: 4:35

Absolutely. So there's clearly a big difference between the, the brick and mortar and traditional grocery business and online grocery shopping, uh, when it comes to personalization and what might be possible. And I say that because in e-commerce every single shopper has the opportunity to have their own unique store. And that's because you don't need to be browsing tens of thousands of unique skews that you're never going to try and never going to buy. And we're going to create a shopping experience that feels like a store designed specifically for you with our grocery partners and in doing so, the recommendations that might be along the lines of reordering things that you've bought in the past, that you're likely running low, um, on your homepage all the way through to a, did you forget type of recommendation engine, right before you check out leads to anywhere from a five to 15% increase in average order value for the grocer. And, uh, that number is really all over the place, depending on geography, depending on shopping frequency. Um, but we play into each of those different metrics differently. And so, uh, taking us out of e-commerce, I'll focus on an in store example. So, uh, uh, fairway markets in New York city, we work with future-proof retail who develops a scan and go mobile checkout technology so that you can actually have that shopping experience in the brick and mortar environment. And we want it to surface, not with everything you scan, but on a couple of products throughout your shopping journey that you're adding directly to your bag. Once you scan them on your phone, so you can pay and skip the checkout line, a couple of recommendations, you've, you've scanned chips, we're going to surface the likeliest salsa. You are to enjoy that. Isn't going to make you walk very far. And over the course of a year, we actually generated a consistent 8% increase in basket size as compared to those shoppers that were not engaging with Hala using the same mobile checkout technology. So both in store and online, there's an opportunity to increase incremental sales via highly individualized shopping experiences.

Phil: 6:40

So it's interesting to me is your fairway example. I'm very familiar with the store and there's a lot of things in fairway that you wouldn't find in a traditional supermarket. Um, you know, it's changing somewhat now that village and, and shop right owns it. Um, but for the most part, you can still find some very unique, uh, New York centric products. So if you look at a Kroger or if you're looking at an Albertson's for the most part, um, what they offer in every store is pretty much the same. I mean, it might have some regional differences, but for the most part, not when you, when you look at a store like fairway, I'm assuming that you had a lot more work to do because of those unique products.

Spencer: 7:31

So if we took a, a traditional approach to recommendations and personalization science, that would be not only right, but I'd be sad about the outcome because it would be again to your point, so much work. Um, the nice thing about how we've developed our engine is that we're really focused on the core commodities, the goods that are being purchased, not the, just the skew level. Cause if you look at product 1, 2, 3 is purchased with product four or five, six, you'll find the best recommendation with anything in any basket, in any grocery store in America is bananas or paper towels. And that isn't super meaningful at the end of the day. So what we're trying to do is to remove the, the noise, if you will involved in this specific skew and to reduce it to it's more abstract entity. So if we've, if there's a specific apple that's sold at a specialty retailer, like a Fairway that isn't sold at most other stores, who've never seen it in our database before what we understand is it's an apple, it's a green apple, it's organic, and here's where it was sourced based on the data that we get provided through the inventory and the enrichments we can add from our proprietary data set, but we don't necessarily need to know how that specific skew has been purchased with others in the past to know that it's an apple and that's the most meaningful part of, of what we do is, is sort of a semantic analysis that, that takes the name of a product it's product ID, a little bit of it's marketing claims, nutrition information, as you pointed out yourself a little while ago, to resolve that to a really basic core good that the consumer understands. And the reason that's most important quite frankly, is because, um, in a big retailer like a Kroger or a Walmart or an Albertson's, what you'll find is they're 2000 different unique rice products, just as an example right now, humans at the best, like the most expert rice, um, chefs in the world are going to really only be able to differentiate between 200 different types of rice. So experts in rice really have the ability to, to, to understand the nuances of the differences between one 10th of all the options out there. So we're trying to figure out the human differentiable attributes of products, not necessarily everything down to how it, the, the amount that it shines differently from one of its competing product. So that layer, that level of, of complexity and granularity is what enables us to say, okay, we're, we're now at the threshold where no one knows the difference between the half a millimeter length and the rice grain. At this point, it's already a saffron basmati or Jasmine rice that's enough for us. And so looking through the human lens rather than the, the merchandisers lens is sort of a key piece to that puzzle.

Phil: 10:28

So everybody is talking about personalization as it relates to grocery shopping. Um, we've, we've heard about it for probably the past five years. How realistic is it, um, in your opinion for customers to really understand the, the value of personalization to the level that you can deliver.

Spencer: 10:54

So it's a really good and important question. And the nice thing is we work with the retailers to reach their end shoppers. And so even if the end shopper doesn't know all the nuance that goes into what's making their experience so much easier, so much more discovery oriented and so much more inspirational, um, the credit should go to the grocery partner that chose to partner with all. And so ultimately we don't have much attribution to the end user. They don't necessarily even need to know, or, or probably don't care in many cases about exactly how hall is doing what it's doing or, or what service or vendor the grocer's using. Rather the grocer is the one that's, that's really focused on those specific questions, because they know that, you know, as cited by BCG in a pretty recent study retail personalization accounts for a 110% increase in likelihood for shoppers to buy unplanned items. And so that's, that's the, um, the piece that Halla's playing into, particularly when we look at online grocery cannibalizing, a bit of brick and mortar sales, you have 20% of your average traditional physical grocery store receipt amounts to, or comes from impulse or incremental sales. So that's not just the stuff in the checkout aisle, that's the scent being pumped from the bakery section, the beautiful array of produce the end cap displays. And so what we're really trying to do is bridge that gap, because what happens to that 20% when you shop online and you don't get those smells, you don't see the produce selection we're using artificial intelligence to say, based on what you've just clicked on, add it to your cart bought last time. It might be running out of these are the likeliest things to compliment what you've you've selected today to make for a slightly bigger basket and a happier customer.

Phil: 12:47

One of the biggest problems that our industry is facing right now is out of stocks. Um, if you take a look at, uh, the latest trend that grocery stores are doing, rather than, as we saw in the beginning of the pandemic, having empty shelves, what they're doing is they're fanning out products. So they're not scaring consumers, uh, but instead of having one SKU of a frozen pizza, they might take seven of them, um, and fill up the space to, to hide the out of stock. What can Halla do for retailer as it relates to out of stock?

Spencer: 13:25

So not only is what you just described, a critical component of the current supply chain and out of stock issues. As we talk about online grocery shopping, but all further that by adding this piece that comes into a stranger's discretion. So when we shop online for groceries, either a store associate or a gig economy worker, who's doing the picking and packing is the one fulfilling your order from a brick and mortar store. And if there's an out of stock item in your order, guess who it becomes up to to decide that substitute that person you've probably never met before fulfilling your order. And they don't know your preferences, they don't know your household. They don't know your cravings. They don't know your dietary restrictions. When you leave it up to their discretion. When the specific spinach you selected was out of stock, they might not know to go for the organic spinach, but instead they'll just grab lettuce. It's another leafy green, and that doesn't leave shoppers very satisfied. So we, we provide is one of our three core solutions alongside recommendations and search. Uh, we power a smart substitution engine, which really enables those pick and pack personal shoppers on their app that they're using to fulfill the order and check each item off a data-driven set of substitutions for each product. That's individualized to the person they're shopping for the user that they're for. So for, for me, that might mean the spinach I clicked was out of stock grabbed me the organic spinach for someone else that might mean get me nothing, right? I wanted that specific spinach. And if it's not there, I don't want it knowing that in advance, rather than needing to interact with the shot, the end customer in real time, ask them questions, send them pictures. We want to make sure we're driving down both cart abandonment from frustrated shoppers and the time spent from the store associate or the personal pick and pack, um, folks that are fulfilling your order. And those two metrics are only possible to be driven down by making this an algorithmic solution, not long that's up to discretion. Okay.

Phil: 15:35

What do you say to that shopper who says Hala knows too much about me, you know, um, I go on Amazon and they pull up recommendations for me all the time. That sort of is, is a bit scary. Um, are you hearing from consumers the same thing that holla, um, maybe knows a little too much. They, they know that, you know, I tend to want organic. So I'm doing that. Um, what's, what's the pushback that if any, that you're getting,

Spencer: 16:10

So this has been a fundamental pillar of how we've designed our platform since day one. Um, Halla does not transact in any personally identifiable information. So even when we work with a retailer, that's got decades of loyalty data to sort of skip over that cold start issue of learning about a shopper. And we can ingest that and have a great understanding from day one of being deployed, we always scramble or ask the, the grocery partner we're working with to anonymize any PII or remove it entirely. So the only thing we actually know about you is your shopper 1, 2, 3, 4, 5, and you shopped here last week, and this is what you bought. And ultimately that what we found is that doesn't make many people uncomfortable at all. It just enhances and enriches the experience. If we knew your name, your address, what you look like, your height, who was in your household, that that's, that becomes invasive. And I got to tell you, I shop on Amazon once in a while. And I went to a concert recently where I had to put in the phone number connected to my Amazon account to enter the, to enter the concert venue, using my p hone's biometric data. That is a level of information that I don't think many people are going to be comfortable with in the long run. And we don't, we've already decided we're not even going to ask for phone numbers, let alone names, let alone p alms. So we're, we're taking a pretty different approach. One that even though the u s doesn't have GDPR or something like it quite yet, we'll play into whatever that future regulation landscape looks like as it pertains to consumer privacy and data. C ause that is a foremost priority for us in running this b est.

Phil: 17:52

So what we're seeing is Tik TOK, Facebook, Instagram, all wanting to incorporate, uh, being able to buy food, uh, from a Tik TOK video by, by just pressing a button. What do you think about that? And do you think that it's something that will become pervasive and the grocers should worry about that?

Spencer: 18:16

So I don't know that grocers should worry about that, even though that's a really good question, and maybe I'm not an expert enough to answer it, but what I can share is that, uh, most of the use cases I've seen with respect to shoppable, social media actually tie back to a grocery retailer. The one fulfilling the order might be, um, an Albertsons Safeway, Kroger, or Walmart giant Eagle, or your regional gross or whatever it might be. And I think that's really important to point out in some cases, yes, it's like a direct to consumer offering from the brand or manufacturer. But that seems to be few and far between relative to I've watching this video, someone's eating Cheetos. I want Cheetos, I click on it and it pops up with three different retailers. I could complete the order from. Um, and I do think it will continue to be pervasive and grow quite substantially, namely due to traceability. Right? The question of return on ad spend has been a huge one for brands and manufacturers for a long time, whether it's about shelf real estate and slotting all the way through to end caps and display ads and signage in store with, with e-commerce tied to social media, the ability to see all the way from how many impressions did this get to how many clicks did it get to actual conversions in specific locations, that level of information on the performance of a campaign around trade spend is something that hasn't been seen before. And I think that visibility is more than valuable enough to fuel huge growth in that direction.

Phil: 19:57

So last question, Spencer, look in your crystal ball. What does grocery look like in five years?

Spencer: 20:07

Well, I think in five years from now, the biggest thing we're going to see is a, a hybrid model of, of grocery shopping. And I say hybrid intentionally, because as much as omni-channel is, is, uh, uh, used in beloved word by many people in this space. There's something about that that makes it sound like it's one seamless altogether experience. I don't think that's going to happen in five years time. Um, and that's not because I don't have faith in technological power and the acceleration of progress in this industry, but rather because it is a simply different experience from any consumer's standpoint, shop online and to shop in store. But I think both will happen. We're going to see a lot of online shopping. Um, I believe play into a replenishment sort of, of realm where things that we know we, we order and use enough that we want replenished or fulfilled on a weekly basis, sort of become an automated process. We will also use online for discovery and inspiration as it pertains to things we bought in the past and our loyalty information at a given retailer. And that will bleed into the in-store shopping environment via mobile engagement, right? That might be QR codes and the like today, or simply looking up a product that's in your shopping list and learning a little bit more about it. Um, some recipes that might be able to be used with, but within five years, I am confident we'll have more AR augmented reality applications as well. That do bring some of the insights that you'd only get an e-commerce experience like, because we know a little bit about your shopping interests and behavior and your household habits, this product that you've just put in the focus of your phone's camera. Now points to these several other alternatives or compliments or recipes you might want to make. And that hybrid digital meets a brick and mortar shopping experience is something that I believe will, will be commonplace in five years.

Phil: 22:08

Spencer, thanks for joining us today on lost in the supermarket. Um, I can't wait to discover more about Halla. I think that you are on the right track. And I think that frankly, uh, every retailer should be talking to

Spencer: 22:25

Thank you so much, Phil. I really appreciate it.