Whether your favorite condiment is Heinz ketchup or your chosen spread on your bagel is Philadelphia cream cheese, ensuring that every one customers have access to their preferred products at the precise place, at the precise price, and at the precise time requires careful supply chain organization and distribution. Amid the proliferation of e-commerce and shifting demand throughout the consumer-packaged goods (CPG) sector, AI and machine learning (ML) have change into helpful tools in enabling efficiency and higher business outcomes.

The journey toward successfully deployed machine learning operations (MLOps) starts with data, says global head of machine learning operations and platforms at Kraft Heinz Company, Jorge Balestra. Curating well-organized and accessible data means enterprises can leverage their data volumes to coach and develop AI and machine learning models. A powerful data strategy lays the inspiration for these AI and machine learning tools to make use of data to detect supply chain disruptions, discover and address cost inefficiencies, and predict demand for products.

“Always remember that data is the fuel, and data, it takes effort, it’s a journey, it never ends, because that is what is absolutely what I’d call what differentiates numerous successful efforts in comparison with unsuccessful ones,” says Balestra.

This is very crucial but difficult throughout the CPG sector where data is usually incomplete given the inconsistent methods for consumer habit tracking amongst different retailers.

He explains, “We do not know exactly and we do not even need to know exactly what persons are doing of their each day lives. What we would like is simply to get enough of the information so we are able to provide the precise product for our consumers.”

To deploy AI and machine learning tools at scale, the Heinz Kraft Company has turned to the flexibleness of the cloud. Using the cloud can allow for much-needed data accessibility while mitigating compute power.  “The agility of the entire thing increases exponentially because what used to take months, now could be done in a matter of seconds via code. So, definitely, I see how all of this explosion around analytics, around AI, is feasible, due to cloud really powering all of those initiatives which can be popping up left, right, and center.” says Balestra.

While it might be difficult to predict future trends in a sector so susceptible to change, Balestra says that preparing for the road ahead means specializing in adaptability and agility.

“Our mission is to thrill people via food. And the technology, AI or what have you ever, is our tool to excel at our mission. Having the ability to learn how one can leverage existing and future [technology] to get the precise product at the precise price, at the precise location is what we’re all about.”

Full Transcript

From MIT Technology Review, I’m Laurel Ruma, and that is Business Lab, the show that helps business leaders make sense of recent technologies coming out of the lab and into the marketplace.

Our topic is machine learning within the food and beverage industry. AI offers opportunities for innovation for patrons and operational efficiencies for workers, but having a knowledge strategy in place to capture these advantages is crucial.

Two words for you: global innovation.

My guest is Jorge Balestra, global head of machine learning operations and platforms at Kraft Heinz Company.

This episode of Business Lab is produced in partnership with Infosys Topaz and Infosys Cobalt.

Welcome, Jorge.

Thanks very much. Glad to be here.

Well, wonderful to have you ever. So persons are likely conversant in Kraft Heinz because it is considered one of the world’s largest food and beverage corporations. Could you check with us about your role at Kraft Heinz and the way machine learning will help consumers within the grocery aisle?

Definitely. My role, I’ll call, has two major focuses in two areas. One in all them is I lead the machine learning engineering operations of the corporate globally. And then again, I provide the entire analytical platforms that the corporate is using also on a worldwide basis. So in role primary in my machine learning engineering and operations, what my team does is we grab all of those models that our community of information scientists which can be working globally are coming up with, and we grabbed them and we strengthened it. Our major mission here is the very first thing we want to do is we want to ensure that that we’re applying engineering practices to make them production ready and so they can scale, they also can run in a cheap manner, and from there we be sure that in my operations hat they’re there when needed.

So numerous these models, because they change into a part of our day-to-day operations, they will include certain specific service level commitments that we want to make, so my team makes sure that we’re delivering on those with the precise expectations. And on my other hand, which is the analytical platforms, is that we do numerous descriptive, predictive, and prescriptive work by way of analytics. The descriptive portion where you are talking about just the regular dashboarding, summarization piece around our data and where the information lives, all of those analytical platforms that the corporate is using are also something that I handle. And with that, you’ll think that I even have a really broad base of shoppers in the corporate each by way of geographies where they’re from a few of our businesses in Asia, all of the approach to North America, but additionally across the organization from marketing to HR and all the pieces in between.

Going into your other query about how machine learning helps our consumers within the grocery aisle, I’ll probably summarize that for a CPG it’s all about having the precise product at the precise price, at the precise location for you. What which means is on the precise product, their machine learning will help numerous our marketing teams, for instance, even once they at the moment are with the most recent generative AI capabilities are showing up like brainstorming and creating recent content to R&D, what we’re attempting to work out what’s the very best formulas for our products, there’s definitely now ML is making inroads in that space, the precise price, all about cost efficiencies throughout from our plans to our distribution centers, ensuring that we’re eliminating waste. Leveraging machine learning capabilities is something that we’re doing across the board from our revenue management, which is the precise price for people to purchase our products.

After which last but not least is the precise location. So we want to ensure that that when our consumers are going into their stores or are buying our products online that the product is there for you and you are going to seek out the product you want, the flavour you want immediately. And so there is a big effort around predicting our demand, organizing our supply chain, our distribution, scheduling our plans to ensure that that we’re producing the precise quantities and delivering them to the precise places so our consumers can find our products.

Well, that definitely is sensible since data does play such a vital role in deploying advanced technologies, especially machine learning. So how does Kraft Heinz make sure the accessibility, quality and security of all of that data at the precise place at the precise time to drive effective machine learning operations or MLOps? Are there specific best practices that you have discovered?

Well, the very best practice that I can probably advise people on is unquestionably data is the fuel of machine learning. So without data, there isn’t any modeling. And data, organizing your data, each the information that you have got internally and externally takes time. Ensuring that it isn’t only accessible and you’re organizing it in a way that you simply do not have a gazillion technologies to cope with is significant, but additionally I’d say the curation of it. That could be a long-term commitment. So I strongly advise anyone that’s listening at once to grasp that your data journey, because it is, is a journey, it doesn’t have an end destination, and likewise it is going to take time.

And the more you’re successful by way of getting all the information that you simply need organized and ensuring that is accessible, the more successful you are going to be leveraging all of that with models in machine learning and great things which can be there to really then accomplish a selected business consequence. So a great metaphor that I prefer to say is there’s numerous researchers, and MIT is understood for its research, however the researchers cannot do anything without the librarians, with all of the those that’s organizing the knowledge around so you possibly can go and truly do what you might want to do, which is on this case research. Always remember that data is the fuel, and data, it takes effort, it’s a journey, it never ends, because that is what is absolutely what I’d call what differentiates numerous successful efforts in comparison with unsuccessful ones.

Getting back to that right place at the precise time mentality, inside the previous few years, the buyer packaged goods, otherwise you mentioned earlier, the CPG sector, has seen such major shifts from changing customer demands to the proliferation of e-commerce channels. So how can AI and machine learning tools help influence business outcomes or improve operational efficiency?

I’ve got two examples that I can say. One is, well, obviously all of us wish to ignore what happened throughout the pandemic, but for us it was a key, very difficult time, because out of nowhere all of our supply chains got disrupted, our consumers needed our products greater than ever because there have been more hunkered down at home. So considered one of the things that I let you know, at the very least for us, that was key was through our modeling, through the information that we have had, we have had some good early warning of certain disruptions in the provision chain and we were capable of at the very least get… Especially when the outbreak began, a few weeks prematurely, we were moving product, we were taking early actions by way of ensuring that we were delivering an increased amount of product that was needed.

And that was because we had the information and we had a few of those models that were alerting us about, “Hey, something is mistaken here, something is occurring with our supply chain, you might want to take motion.” And taking motion at the precise time, it’s key by way of getting ahead of numerous the things that may occur. And in our case, obviously we live in a competitive world, so taking actions before competition is significant, that timing component. One other example I can provide you with and is more of something that’s we’re doing increasingly nowadays is that this piece that I used to be referring to about the precise location about product availability is vital for CPG, and that’s measured in something that is named the CFR, and is the client field rate, which implies is when someone is ordering product from Kraft Heinz that we’re capable of fulfill that order to 100%, and we predict to be really high with high 90s by way of how efficient we’re filling those orders.

We have now developed recent technology that I feel we’re pretty pleased with because I feel it is exclusive inside CPG that permits us to essentially predict what will occur with CFR in the longer term based on the precise actions we’re taking today, whereas it’s changing our production lines, whereas changes in distribution, et cetera, we’re capable of see not only the immediate effect, but what is going on to occur in the longer term with that CFR so we are able to really act on it and deliver actions at once which can be in the advantage of our distribution in the longer term. So those are, I’d call it, say, two examples by way of how we’re leveraging AI and machine learning tools in our day-to-day operations.

Are those examples, the CFR in addition to the provision chain and ensuring consumers had all the pieces on demand almost, is that this unique to the food and beverage industry? And what are perhaps another unique challenges that the food and beverage industry faces once you’re implementing AI and machine learning innovations? And the way do you navigate challenges like that?

Yeah, I feel something that may be very unique for us is that we all the time should cope with an incomplete picture by way of the information that we’ve in our consumers. So in the event you give it some thought, once you go right into a food market, a few things, well, you’re buying from that store, the Kroger’s, Walmart’s, et cetera, and a few of those could have you identified by way of what’s your consumption patterns, some is not going to. But in addition, in our case, in the event you are going to go buy a Philadelphia [cream cheese], for instance, you might decide to buy your Philadelphia in multiple outlets. Sometimes you wish more and also you go to Costco, sometimes you wish less, in my case, I live within the Chicago land area, I am going to a Jewel supermarket.

We all the time should cope with incomplete data on our customers, and that could be a challenge because what we try to work out is how one can higher serve our consumers based on what product you want, where you are buying them, what’s the precise price point for you, but we’re all the time coping with data that’s incomplete. So on this case, having a transparent data strategy around what we’ve there and a transparent understanding of the markets that we’ve on the market so we are able to really grab that incomplete data that we’ve on the market and still provide you with the precise actions by way of what are the precise products to place, just to present you an example, a transparent example of it’s… And I’m going back to Philadelphia because, by the way in which, that is my favorite Kraft product ever…

Philadelphia cream cheese, right?

Yes, absolutely. It’s followed by a detailed second with our ketchup. I even have a soft spot for Philadelphia, pun intended.

– and the ketchup.

Exactly. No, but you have got different presentations. You may have the spreadable, you have got the brick of cream cheese, throughout the brick you have got some flavors, and what we would like to do is ensure that that we’re providing the flavors that individuals actually need, not producing those that individuals don’t desire, because that is just waste, without knowing specifically who’s buying on the opposite side and you ought to buy it in a supermarket, one or two, or sometimes you’re shifting. But those are the things that we’re consistently looking out for, and clearly coping with the truth about, hey, data goes to be incomplete. We do not know exactly and we do not even need to know exactly what persons are doing of their each day lives. What we would like is simply to get enough of the information so we are able to provide the precise product for our consumers.

And an example like cream cheese and ketchup probably, especially if a child is in the home, it’s considered one of those products that you simply use on a reasonably each day basis. So knowing all of this, how does Kraft Heinz prepare data for AI projects, because that in itself is a project? So what are the primary steps to prepare for AI?

One thing that we’ve been pretty successful on is what I’d call the potluck approach for data. Meaning that individual projects, individual groups are focused on delivering a really specific use case, and that’s the precise thing to do. If you end up coping with a project in supply chain and also you’re trying simply to, for instance, say, “Hey, I would like to optimize my CFR,” you’re really not going to be caring that much about what sales desires to do. Nonetheless, in the event you implement a potluck approach, meaning that, okay, you wish data from someone else, and it’s totally likely that you have got data to supply because that is a part of your use case. So the potluck approach implies that if you ought to check out the food of someone else, you might want to bring your personal to the table. So in the event you do this, what starts happening is your data, your enterprise data, becomes little by little more accessible, and in the event you do it right eventually you just about have lots and almost all the pieces in there.

That’s one thing that I’ll strongly advise people to do. Think big, think strategically, but act tactically, act knowing that individual projects, they will have more limited scope, but in the event you establish certain practices around sharing around how data needs to be managed, then each individual projects are going to be contributing to the larger strategy without the biggest strategy being a burden for the person projects, if that is sensible.

Sure.

So at the very least for us that has been pretty successful over time. So we’ve data challenges absolutely as everybody else has, but at the very least from what I have been capable of hear from other people, but Kraft Heinz is in a great place by way of that availability. Because when you reach a certain critical mass, what finally ends up happening is there is no must bring additional data, you’re all the time now reusing it because data is large however it’s finite. So it isn’t infinite. It isn’t something that is going to grow ceaselessly. Should you do it right, you must see that eventually, you do not need to herald increasingly data. You simply must fine-tune and really leverage the information that you have got, probably be more granular, and possibly get it faster. That is a great signal. I even have the information, but I would like it faster because I would like to act on it. Great, you are on the precise track. And likewise your associated cost around data should reflect that. It shouldn’t grow to infinity. Data is large but is finite.

So speaking of getting data quickly and making use of it, how does Kraft Heinz use compute power and the cloud scaling ability for AI projects? How do you see these two strategies coming together?

Definitely the technology has come a good distance in the previous few years, because what cloud is offering is more of that flexibility, and it’s removing numerous the constraints, each by way of the dimensions and performance we used to have. So to present you an example, just a few years back I needed to worry about “Do I even have enough storage in my servers to host all the information that we’re getting in?” After which if I didn’t, how long is it going to take for me so as to add one other server? With the cloud as an enabler, that is now not a problem. It’s just a few lines of code and also you get what you wish. Also, especially on the information side, a few of the more modern technologies, talking about Snowflake or BigQuery, enable you to separate your compute out of your storage. What it principally means in practical terms is you do not have people fighting over limited compute power.

So data could be the identical for everybody and everybody could be accessing the information without having to overlap one another after which fighting about, oh, in the event you run this, I cannot run that, after which we’ve all varieties of problems so definitely what the cloud allowed us to do is get out of the way in which by way of the technology as a limitation. And the good thing that happened down there now with all of the AI projects is now you would give attention to actually delivering on the use cases that you have got without having to have limitations around “how am I going to scale?”. That isn’t any longer the case. You may have to fret about costs since it could cost you an arm and a leg, but not necessarily around how one can scale and the way long it is going to take you to scale.

The agility of the entire thing increases exponentially because what used to take months, now could be done in a matter of seconds via code. So definitely I see how all of this explosion around analytics, around AI is feasible, due to cloud really powering all of those initiatives which can be popping up left, right, and center.

And speaking about this, you possibly can’t really go it alone, so how do partners like Infosys help herald those recent skills and industry know-how to assist construct the general digital strategy for data, AI, cloud, and whatever comes next?

Much in the identical way that I feel cloud has been an enabler by way of this, I feel corporations and partners like Infosys are also that sort of enablers, because, in a way, they’re a part of what I’d call an expertise ecosystem. I do not think any company nowadays can do any of this by itself. You would like partners. You would like partners that each are bringing in recent ideas, recent technology, but additionally they’re bringing in the precise level of experience by way of those that you wish, and in a worldwide sense, at the very least for us, having someone that has a worldwide footprint is significant because we’re a worldwide company. So I’ll say that it is the same thing that we talked about earlier about cloud being an enabler: that expert ecosystem represented by corporations like Infosys is just one other key enabler without which you’ll really struggle to deliver. So that is what I’ll probably say to anyone that’s listening at once, ensure that that your ecosystem, your expert ecosystem is nice and is flourishing and you have got the precise partners for the precise job.

When you consider the longer term and likewise all these tough problems that you simply’re tackling at Kraft Heinz, how essential will something like synthetic data be to your data strategy and business strategy as well? What’s synthetic data? After which what are a few of those challenges related to using it to fill within the gaps for real-world data?

In our case, we do not use numerous synthetic data nowadays because at the very least from the areas that we’ve holes to fill by way of data is something that we have been coping with for some time. So we’re, let’s put it this manner, already familiar on how one can produce and fill within the gaps using a few of the synthetic data techniques, but not likely to the identical extent as other organizations are. So we’re still searching for opportunities when that’s the case by way of what we want to make use of and leverage synthetic data, however it’s not something that least for Kraft Heinz and CPG in any respect we use extensively in multiple places as other organizations are.

And so, lastly, once you think ahead to the longer term, what’s going to the digital operating model for an AI-first firm that is focused on data appear to be? What do you see for the longer term?

What I see for the longer term is, well, initially, uncertainty, meaning that I do not think we are able to predict exactly what is going on to occur because the realm specifically is growing and evolving at a speed that I feel is just truthfully dazzling simply because of the most important things. I feel at the very least what I’d say is the actual muscle that we have to be exercising and be ready for is adaptability. Meaning that we are able to learn, we are able to react, and apply the entire recent things which can be coming in hopefully at the identical speed that they are occurring and really leveraging recent opportunities once they present themselves in an agile way. But at the very least from the how one can prepare for it I feel it’s more about preparing the organization, your team, to be ready for that, really act on it, and be ready also to grasp the precise business challenges which can be there, and search for opportunities where any of the brand new things or perhaps existences which can be happening could be applied to resolve a selected problem.

We’re a CPG company, and which means the precise product, right price, right location, so anything boils right down to how can I be higher in those three dimensions leveraging whatever is accessible today, whatever’s going to be available tomorrow. But keep specializing in, at the very least for us, we’re a CPG company, we manufacture in Philadelphia, we manufacture ketchup, we feed people. Our mission is to thrill people via food. And the technology, AI or what have you ever, is our tool to excel at our mission. Having the ability to learn how one can leverage existing and future to get the precise product at the precise price at the precise location is what we’re all about.

That is incredible. Thanks a lot, Jorge. I appreciate you being with us today on the Business Lab.

Thanks very much. Thanks for inviting me.

That was Jorge Balestra, global head of machine learning operations and platforms at Kraft Heinz Company, who I spoke with from Cambridge, Massachusetts, the house of MIT and MIT Technology Review.

That is it for this episode of Business Lab. I’m your host, Laurel Ruma, I’m the director of insights, the custom publishing division of MIT Technology Review. We were founded in 1899 on the Massachusetts Institute of Technology, and you may also find us in print, on the internet, and at events every year around the globe. For more details about us and the show, please take a look at our website at technologyreview.com.

 This show is accessible wherever you get your podcasts. Should you enjoyed this episode, we hope you will take a moment to rate and review us. Business Lab is a production of MIT Technology Review. This episode was produced by Giro Studios. Thanks for listening.