Retail and CPG: Collaborating on Data Drives Growth at Scale

Summary

Walmart and Unilever replaced traditional batch EDI and API-based data sharing with Databricks Cloud Feeds, enabling direct cloud-to-cloud access to 65+ real-time datasets including POS, demand signals, and inventory data, with unified governance through Unity Catalog.
The collaboration delivered 5.3% year-over-year growth, 98% in-stock rates, and $1.5 billion in annual sales, driven by AI models that continuously optimize forecasting, replenishment, and inventory using granular, real-time retail signals.
Cloud Feeds eliminates API overhead and enables higher data atomicity, allowing Unilever to build perfect store programs, target specific markets with geo-location insights, and execute revenue growth management decisions that previously depended on day-old data.

Retail and CPG: Collaborating on Data Drives Growth at Scale

Supply chain volatility is the new normal. Traditional siloed forecasting, batch EDI, and reactive logistics cannot keep pace. Walmart and Unilever transformed their partnership by replacing linear supply chains with intelligent networks that sense and respond in real time. By unifying retail POS, demand signals, and inventory data through Databricks open sharing and Cloud Feeds, they power AI models that continuously optimize forecasting, replenishment, and inventory. Their results: 5.3 percent year-over-year growth, 98 percent in-stock rates, and 1.5 billion dollars in annual sales.
Learn how Cloud Feeds eliminates API overhead through direct cloud-to-cloud data access with unified governance via Unity Catalog. Discover the principles enabling collaboration: tech-agnostic architecture, skill-inclusive design, and governed distribution. See how real-time data from Walmart's 65+ data sets enables Unilever to build perfect store programs, target specific markets with geo-location insights, and execute revenue growth management at scale. The result: a blueprint for CPG-retail partnerships that turn data into competitive advantage.
🤝

Chapters

FAQs

What is Databricks Cloud Feeds and how does it differ from traditional API-based data sharing?

Cloud Feeds, formerly known as Open Sharing or Delta Sharing, enables direct cloud-to-cloud data access without API overhead or data extraction steps. Unlike APIs that require data to be pulled and delivered on a schedule, Cloud Feeds lets the consumer access data directly in the provider's cloud storage, enabling higher granularity and lower latency.

What data does Walmart share with Unilever through the Syntela platform?

Walmart's Syntela platform provides Unilever access to 65+ datasets covering shopper behavior, retail performance, and market perception data. This real-time, granular data allows Unilever to build AI models for demand forecasting, perfect store programs, and targeted revenue growth management.

What business results did the Walmart and Unilever data collaboration achieve?

The collaboration delivered 5.3% year-over-year growth, 98% in-stock rates, and $1.5 billion in annual sales. These results came from AI models using real-time retail data to continuously optimize forecasting, replenishment, and inventory allocation across Unilever's portfolio.

Why does lower data latency matter for CPG companies like Unilever?

Traditional batch approaches delivered retail data 24 to 48 hours after the fact, which is too late to act on out-of-stock events or trending demand signals. Databricks Cloud Feeds reduces latency to near-real time, enabling Unilever to respond to inventory gaps, time targeted advertising, and execute market-specific programs like geo-targeted campaigns with precision.

Full transcript

[00:08] First of all, I thank you for being here. Thank you for the speakers that can come up from Walmart, from Unilever. Thank you for Walmart and for Unilever for helping to forge the ground that we're going to talk about today. This has been a passion of mine um at Databricks. I'm Rob Saker. I lead the global consumer
[00:24] industries here. And you know, think about the challenges that we have in the industry right now. We need to get get to profitable volume growth. We have a challenge of being very price dependent. And the way that we're going to get there is by taking the inefficiencies out
[00:39] and reinvesting those areas that we can actually hopefully drive more profitable activities. I think what you're going to see today is really interesting because our biggest enemy in getting there is time. And if you think about when I used to
[00:54] have a team of tens of thousands of people that would go out and do on-shelf availability or price audits or other things there, how did I get the information to do this? I would extract the information. The retailer would extract the information. They would send it over to me at 12:30 at night.
[01:10] I would take it, bring it in, process it. Maybe a day or two later, I would have it. I'd be running my predictions against it. I'm delayed a couple days. And that latency is that's the time of the opportunity inside of unresolved out of stocks.
[01:25] That's just an easy example of where I missed the opportunity. So, what we've been pushing, what used to be called Delta Sharing, is now called Open Sharing, radically changes that. It's not just about making it faster, though.
[01:41] When you make it faster, you're actually able to make it more granular, more atomic. You're able to go to a higher level of atomicity in the information. And then it opens a lot of interesting doors. And what Walmart is really forging the way in the industry in doing this by moving to the Open sharing
[01:57] protocol allows them to get to really great levels atomicity that open up a whole set of new use cases for targeting behaviors, when to run ads, when to do certain behaviors and more. Uh so, I'm really thrilled to have Walmart and Unilever come up and talk
[02:12] about the experience they've been going on this journey. Uh and um you know, with that, we're going to have first Walmart. So, Karen, you want to come up here and uh walk through the As she's coming up, the what I'd say is um we'll have time at the end if you want to have questions. No questions throughout, but we'll stick around at
[02:29] the front um and then happy to take questions that you may have after the session. Okay, great. All yours. Usually, I don't need a mic, but I'm going to try I'm going to try not to um talk into your ears too loud.
[02:44] Um all right. So, um So, Syntela um is a product that provides um information about various journeys of of a Walmart shopper. Uh we have modules like shopper behavior that
[03:01] uh lets our suppliers and partners understand uh how a shopper is is is shopping. You know, what is the basket mix mix, you know, what are they buying, what is the item affinity. If somebody buys a tooth toothpaste, do they also buy a toothbrush or floss and things of
[03:18] that nature. Um taking a step for further, how how does a shopper actually reach uh what they're trying to buy is provided through digital landscape, you know, all the way from what keywords do they enter and then what actually gets them to the
[03:33] product that they're they're looking for. We combine all of this information through insights activation that that makes advertisers life easier to use all of these these data sets to then target ads uh to the to the final customer.
[03:50] Then uh we have customer perception uh where we have the largest panel of real Walmart shoppers where our suppliers can publish questionnaire or, you know, sort of you know, if they want to get feedback on an idea that
[04:06] they have about launching random things like, for example, some of you decided to launch a colored toilet paper. You know, is that if that's a good idea or not, you can find that out through customer perception by sending a a survey uh to real Walmart shoppers. Most
[04:22] likely, it will be a bad idea. Then, there's channel performance. As the name suggests, we provide various operational metrics through these data sets about sales and other fulfillment metrics um that provide various
[04:40] information about supply chain and operational efficiency around that. So, I'm going to skip this slide because we just talked about channel performance. It just kind of talks about what sort of data sets that we have. To really talking about what is the
[04:56] challenge at hand at Walmart when we talk about supplying these data sets to the end consumer. So, let's try to understand the scale, right? Store sales is a data set that we provide as a part of our offering.
[05:11] And one store sales data set for a CPG customer or CPG supplier, you're looking at 3,830 items. And that's an average number. It keeps going up, keeps going down across all the service channels at
[05:27] Walmart. Um 4,600 stores in US. We're international. The data sets that we provide is also for international market. So, that's why I've noted that it's only for US. Um we provide 2 years of historical data and about 30 KPIs just for store sales.
[05:45] This alone is 5.5 gigs every day for one one supplier. Now, this is one out of 65 data sets that we deliver every day to a supplier. So, think about that scale and you know, what we'd need to
[06:01] deliver all of this information to each and every supplier that we have at Walmart. So, 6 years ago, we embarked on a journey to create a product offering that the suppliers and partners can access
[06:18] programmatically to power their models, their applications, and and things of that nature. So, we put together a So, at the time we obviously did not have 65 data sets, but now we do have 65 plus robust data sets
[06:34] from various operating systems around sales, inventory, fulfillment, and things of that nature. So, we we have a foundation where we bring all of these data sets together, and then we we built two API endpoints for each of these
[06:50] each of these data sets. One to figure out whether the data set is ready or not because the frequencies are different. So, you can pull that and try to understand if the data set is available or not, and one endpoint to actually download the data. So, what that meant was in this cycle,
[07:07] you're continuously calling an endpoint trying to understand if a data set is available or not, and when it is, you call the other endpoint and download that data. And then the once you've retrieved that information, the journey to do an ETL and all sorts
[07:24] of processing on it suddenly, okay. Start, and then you store the data, and then you make it available to the your end consumer for analysis and reporting. This introduced some overhead uh to the
[07:40] consumer because uh, of course it requires certain level of skill sets uh, to uh, to be had with the with the supplier and partners. So, you know, it requires some level of API development, some level of data engineering and things of that nature.
[07:56] Um, and at the same time infrastructure as well where you need storage and compute. Um, so we wanted to make sure that as we are evolving the product for the next generation of use cases particularly AI in today's world uh,
[08:13] we're making it easier for the suppliers and partners to consume this information. And so as we embarked on the journey to to uh, to kind of evolve this, we had some guiding principles. We wanted the
[08:28] solution to be tech agnostic. Uh, I know this is all Databricks. Uh, we love Databricks, but we want as as Walmart Data Ventures, we wanted to make sure that we're not pushing anybody in a particular direction. So, any solution that we build has to be tech agnostic.
[08:45] No matter where you are, we'll meet you there. Skill inclusive. So, we wanted to we did not want to introduce more uh, skill overhead to our consumers. So, we wanted to make sure that, you know, they can consume data and products that we're building uh, with the skill sets uh,
[09:00] that they have with a minimum amount of uplift there. At at Walmart, you have to have governance first. So, we want to make sure that that is also uh, on the top of our priority list. And then all of these complexity we needed to create under a an abstraction layer. So, so we did all
[09:17] of that and we finally launched what we call as cloud feeds which is based on Databricks open share uh, protocol. Um, what that means is you're still looking at those 65 plus robust data
[09:33] sets but you're not going through an API, you're not pulling it, uh you're not trying to understand its status through those endpoints, um and then downloading the data, and then doing ETL on it, and then finally making and then after all of these
[09:48] processing, making it available to your your end consumer. What this does is uh it it activates uh direct cloud-to-cloud sharing where these data sets are sitting in Walmart cloud in in a particular format that is
[10:04] required by Open Share. We used to call it Delta Share, it's now called Open Share. And we will provide you with a dot share file which has all the credentials, has all the partition details, and you upload that into your cloud, uh and then
[10:20] you get direct access to these tables without having to call an API, uh download the data, carry on your ETL process, and things of that nature. You could be on any cloud platform, or you could your applications could be programmed in any language of your
[10:35] choice. Um you get direct access to to to uh to the data sets that you have subscribed to. So, like I said, uh the benefits of cloud feeds are the fact that you get direct access to data. Uh most importantly, and you you'll hear
[10:51] our colleagues from Unilever talk about it, it reduces uh your operational overhead, your compute, storage, skill set required, all all of those those things. Um and one thing that is really important to understand is um especially in our
[11:07] business, there are a lot of data restatements that take place. You know, you could have audit in inventory, you could have, you know, things that will change data that you have already consumed. Um and in a and I I I I So, I actually did have a slide but
[11:23] which is missing here. But then uh um in historical corrections context, what you need to do is we need to create these files again. And uh um you'll need to download the files again and then carry out the the whole ETL process again.
[11:39] In this scenario, these these corrections are directly made on the table. And that your programs are already consuming, so you don't need to do anything. You just need to continue to read uh those tables directly. Uh I'm happy to inform that we've also
[11:54] reached 100% parity with the API uh data feeds product that we had before. So, any customer or any consumer of Sentila cloud feeds can can without a doubt move to uh migrate to cloud feeds from APIs.
[12:11] Uh there's you know, you'll get access to all the data sets which is listed out here. Um also, an important thing to understand is as we've built this product on open share and we've built the Unity Catalog on top of it.
[12:29] It actually readies us for collaboration in the future. So, think about we've already talked about how we made it easier for the consumers to access data directly without with minimum movement of data. So, you're store you're you're
[12:44] saving on storage and compute. Uh with this, you can also get the data delivered to whichever cloud uh you're using. And ultimately, as we built the Unity Catalog, it is also making sure that we
[13:00] enable some of the future data collaboration capabilities where you can have multi-party collaboration activated. So, you can bring more people to collaborate with you with the data that uh that you have access to to um
[13:15] to generate insights that going to that's going to unlock more use cases such as in ads. Um but uh but that you know, it becomes it becomes really easy to to to bring more collaborators to bring their own data
[13:32] sets that could be sensitive. But in a in a secure collaboration uh environment. With that, I'm going to hand it over to uh my colleagues from Unilever. Thank you.
[13:53] Just in here. All right. Everyone, how's everybody doing? One more round of applause for Kieran. I I think we need to change sides so that we represent the flag appropriately. There we go. We got stars and stripes here. Um
[14:09] we we will introduce ourselves twice uh it looks like. Uh thank you Kieran. Thank you Rob for letting us share the stage with you. Um Mike and I are from from Unilever. I'm Kevin Casey. I'm Cooper. How are you? Uh we we work at at Unilever uh in a
[14:25] group called customer analytics. Uh I I have the pleasure of leading the customer analytics team. We work with all of our retail partners in the US. Um and our single focus and priority is to use all of the data and all of the signals coming from our retailer
[14:40] partners, coming from third parties, you think about syndicated data sources, and others to drive growth for Unilever and our retailers. And um and we've got a pretty exciting story today that that's um we're going to share with you guys about how we've we've utilized Cloud Feeds uh
[14:57] to do just that. Um if you're not familiar with Unilever, I I hope everyone's familiar with Walmart, but if you're not familiar with Unilever, we're a global company. We operate in over 185 countries around the globe. Uh we're based out of Europe, so you see that we have 50 billion dollars
[15:13] or 50 billion euros uh not dollars in in annual turnover. And uh we're proud to say that our our products serve uh 3.7 billion people every single day. You may not recognize this as a company, but hopefully you recognize our brands
[15:29] and and and those brands hopefully you guys are some of those 3.7 billion that know and love our brands like Dove. It says Rexona here if you're if you're from the United States, that might be Degree deodorant. You might recognize that one. Liquid I.V. if anyone's been out late at
[15:45] the at the event last night, hopefully you got some Liquid I.V. this morning. Um and if you I'm not sure what they put on the sandwiches, but maybe there was Hellmann's mayonnaise on that. Um so our products show up. We're not a D2C
[16:01] company. Uh when we talk about our customers, we're talking about Walmart. We're talking about you know, the other retailers that we might go and and buy these products in stores. Um so hopefully you guys will will shop for these brands at Walmart.
[16:18] Um I have the pleasure of looking at you know, working on the the whole all customers and and what we call customers being Walmart. All customers in the US, but I did work on the Walmart team about 6 years ago and and obviously you've heard from Kieran Walmart data sharing
[16:34] has evolved. When I worked on on Walmart 6 years ago, uh it looked like this. Are there any CPG folks in the room? Anybody from CPG? I you probably recognize that this is the old front end to to Retail Link. Uh
[16:50] I left the Walmart business and and when I when I came back All it took was for you to leave the business for the evolution to start. Mike, what what's what's the deal now? So Kieran has done a great job talking about Sintelix from the Walmart perspective. We wanted to share just a little bit about Sintelix from from the
[17:06] vendor or supplier perspective and how how game-changing this has really been for us. So quick survey. Who here has been to a Walmart store in the past year. Raise your hands. Okay, if you did not raise your hand,
[17:22] you're one of the 5% of Americans who did not shop a Walmart store in the past 12 months. Now, think about that. 95% of Americans are shopping in Walmart at least once or twice during the course of the year. Globally, it's a business that's going
[17:38] to do about $725 billion, right? Like the mass size of what Walmart is. And depending on the category, that 25% of that is going to be done online. So, they are a true omni retailer. They've got brick and mortar and all the
[17:54] channels online that you that you need to fulfill um your shopping list week to week, month to month, right? So, just just massive. And they are for Unilever, they are a big piece of our business globally, right? If if Unilever with
[18:10] Walmart I should say it differently. If Walmart were a country for Unilever, it would they would be the fifth largest country in the globe. They are a massive partner for Unilever. We want to make sure that we are serving them as best we possibly can. And the key unlock to that is data.
[18:29] Okay. So, again, as Kevin talked about some of the um the cloud feeds and the transition from APIs and Retail Link to the new Sintel of the cloud feed platform, it has opened up so much for us to be able to do um with the data and how we
[18:46] go to market with them. And so, we're going to do Unilever is going to do about $3 annual with Walmart. It's a again, it's a massive chunk of business for us. And what we wanted to do through this platform is really make
[19:03] sure through the new ways of of acquiring data from Walmart is making sure that we're maximizing every bit of efficiency. You know, Kevin talked about it a little bit. They Walmart wants this data to permeate the organization, not just the account team that serves Walmart, the organization. They want us
[19:19] to embed this in our processes. They give us more data than than nearly any other customer. And where Karen talked mostly about the the channel performance data set, the shopper behavior, the fulfillment data, supply chain data, and all these other
[19:35] data sets are part of that system for us as well to really make sure that we're honing in and taking advantage of every nugget every piece of visible data that we possibly can to maximize our POS and and performance at Walmart. Okay.
[19:51] We just to wrap up this slide the the collaboration that we that this this data set is sparking within Unilever is tremendous. It goes everywhere. It it informs so many decisions for us. Again,
[20:07] as Walmart is a number one partner for Unilever in the US, we really are are struggling or not struggling, we're really trying to maximize how that data impacts our POS. Okay. Awesome. Switching gears just a little bit. I you guys you see the stars and stripes.
[20:24] Uh what's all that about? Has anyone heard of a thing called the FIFA World Cup that's happening right now? Any any soccer fans in the room? Okay. Um I I'm going to go out on the record. This is being recorded. United States is going to win the World Cup, right?
[20:39] Right? No? Anybody else? Thank you. Uh Spain? Anybody got Spain? Okay. Brazil? Any Anybody got Brazil? Um England? There it is. Um we as Unilever are proud to be
[20:57] partners of the FIFA World Cup. And we thought that we'd bring that to life a little bit for you guys today, not just with the uh the the love for the United States, Uh with the story about how we're utilizing the data we're getting from Walmart with a real-world application that's relevant right now. So, Mike, what are
[21:15] you guys doing uh with FIFA? Okay, FIFA. It's a fantastic event. It is a number one global event every 4 years, second only to the Olympics, I believe. All right. So, a really really important moment for Unilever. And we are Unilever is a global partner
[21:32] with with FIFA on our personal care categories. So, we want to maximize every bit of potential with this with this event that we possibly can. Unilever has several different platforms and things in motion right now as I'm
[21:47] sure most CPG companies do. One of those things for us is something we like to call perfect store. So, what's the definition of perfect store? We want to make sure that we've got the right inventory, that we've got the right shelf, that we've got the right display, that we've got the right service, that we've got the right
[22:02] inventory. Everything is perfect so that when a shopper goes into a Walmart store, they can find that thing that they are looking for with ease, put it in their basket, and convert that sale for us, right? That's a bene- It's a huge benefit to both Unilever and to
[22:19] Walmart. We want to maximize that as as best we possibly can. So, again, we want to be unmissable. We want to be in the right stores. We want to make sure that that it's the online store as well as the brick-and-mortar store, and that it's always available through
[22:34] OSI. through inventory. Okay. So, as we thought about this program in in terms of our perfect store, how are we going to bring this to life with Walmart? What are the things that we are going to have to do? Inventory management is kind of where we
[22:50] started out. Do we have the right supply chain going on? Do we have the supply to feed Walmart based off of the information that we get through again through the Sensormatic data? The fulfillment process. Do we have enough inventory in stores so
[23:05] that Walmart can can fulfill those orders? And how and make sure that their consumers or their customers are shopping or getting the things that they're looking for and shopping for. And that's just one side of it. We also we look around at other disparate data sources that we have access to. It's it
[23:22] mostly is about the Walmart data, but we have to accentuate that with other sets of data. So, we start looking at geo-locational data like where people say come to the US and participate in in FIFA, where those events going to be? Where are they going to be staying? How long are you going to be here? We do a
[23:39] deep dive on that. We we pick up short-term rental information as well. Which are the biggest cities? How much How many people are expected to travel to Dallas or or San Francisco or New York? How long are they going to be staying for? Are they going to have to pick up personal care items while they're in the United States?
[23:56] And if you do just a little bit of research, you'll you'll understand that the average traveler to the US for FIFA, they're going to spend about seven or eight days in the United States and they're going to spend over $400 each in shopping while they're here.
[24:12] So, let's take advantage of that moment. They're going to be in the stores. They're going to They're going to be looking for items. Let's make sure that they're um that they're bringing their dollars to Unilever and to Walmart. Okay.
[24:29] Kevin? Mike, it's uh incredible work and uh and the good news is the business impact is clear. Um we talk you know, Kieran talked about the ease of shifting from uh from API to cloud feeds. Uh we we're not looking back. Um
[24:45] We are not. Cuz cuz Mike mentioned Walmart wants uh this is this is a huge mindset shift from those retailing days where data was was a little bit harder to access and and kind of you know, isolated in the in your field sales teams, or at least our field sales teams. And the mindset shift
[25:00] that they want that data to permeate the organization. And the and cloud feeds have uh have certainly unlocked that. It's made it so much easier um for us to collaborate, not just with our our Walmart partners, but with our internal teams as well. Mike and I sit
[25:16] in the field. Mike sits in uh in Northwest Arkansas uh with our sales team. And the data flowing in on a real-time basis is critical for us to make decisions in real time. But also to bring those insights back to our internal teams and ensure that they know
[25:32] what it takes to win with Walmart and the Walmart consumer. And and the data that Walmart's uh making available to us allows us to do that with speed. And allows us to get to those insights um so much faster than uh than I was there 6 years ago. So, pretty exciting stuff.
[25:48] We uh we don't do it all on our own. Uh Mike has an incredible team. Uh but we also work with third-parties. We have a lot of tools and capabilities that we're building internally that we leverage uh external partners uh to build on our behalf. And being able to feed that data and have very granular uh data available
[26:06] allows us to build uh really really cool analytics. Um and do that that rapid technology advancement, build those next generation uh capabilities around perfect stores as Mike mentioned, around you know, think about concepts like revenue growth
[26:22] management, forecasting, the granularity, the speed which the data comes through uh is really incredible. And we we really believe that Walmart is uh is setting the standard these days for the for the retail industry. Um and it's not just, you know, Mike mentioned some of the the benefits of, you know,
[26:39] you know, it's faster, it's easier to manage, it's the the cost savings. But but that that's not what what we're uh we're tasked with. We're tasked with growth. And we're excited to say that year-to-date at Walmart, uh Unilever has has done 1.5 billion dollars of sales.
[26:57] That's a 5.3% year-over-year growth rate, which is phenomenal. We're growing market share. And we have alignment with with Walmart's leadership team to grow 10% in the in 2027 and beyond
[27:13] because of the partnership that we've been able to establish. Mike, you know, with the perfect store program, there's a huge focus on the right items in the right stores at the right time or online. Whatever the shopper wants to to make or have our products in their
[27:30] baskets, we have to make sure that that's that's available. And proud to say you guys have achieved a 98% in-stock rate this year, which which is industry-leading and and definitely beats the benchmarks that that Walmart would would ask us to achieve.
[27:45] Um, so with that, we again want to want to say thank you to to Kieran and, you know, Walmart for for the partnership. Databricks obviously makes this available to us or we couldn't do it without
[28:01] technologies like Databricks. We're super excited about all the things that we've heard this week. All the new advancements, all the new technologies that's that's coming out. You can't do any of that stuff without the data, as we all know. That's not the problem anymore. We get to focus now on doing
[28:17] the really cool analytics, taking advantage of the new technologies, and focusing our time and energy there. So, that's what we had to share with you guys today.

Learn more about the Databricks Data and AI platform.

The information provided herein is for general informational purposes only and may not reflect the most current product capabilities or configurations.