Setting a New Bar for a New Supply Chain

The full transcript and presentation slides from the thought provoking conversation are below.

Peter Breen: Welcome to “Building a Supply Chain for Omnichannel Commerce.” I'm Peter Breen the editor in chief of the Path to Purchase Institute and I'll be your moderator today. On behalf of Consumer Goods Technology, I'd like to welcome you.

The key areas of focus for the discussion are manufacturing, fulfillment, that last piece of fulfillment that everybody's calling the last mile, and then finally demand planning, which, of course, has a major impact on those other three areas.

We are delighted to have a couple of subject matter experts to help us talk through the topic. Fred Fontes is a client partner with Aera Technology, where he helps customers drive impact across the enterprise by digitizing and augmenting complex decision-making. Before joining Aera, he was an associated partner in McKinsey’s advanced analytics practice, where he led clients through large-scale analytics transformations across CPG and retail.

Arnaud Morvan is Aera's senior engagement director for customer engagement, where he applies nearly 20 years of end-to-end supply chain experience to help clients undertake digital transformation. His experience before Aera includes positions at retailer Carter’s, at consumer productions manufacturer Hilti, and with analytics company Symphony Retail Solutions.

Before we dive into covering some of what was in the report, I wanted to get things started by asking one general question, and if I can pose this to you, Fred: What has been the single greatest cause of disruption to the traditional supply chain that you've seen in the last five years?

Fred Fontes: I love that we're kicking off with an easy question here, Pete. I think especially over the last 90 to 100 days, we'd be hard-pressed to not be able to point to a myriad of things that have impacted global supply chain in massive ways. Whether we just looked around the corner with what's happening with coronavirus in China and how that's impacting supply chains. Or Brexit and how that's disrupting supply chains in Europe.

But I keep on coming back to what I think is probably, especially in the consumer space, one of the biggest disruptions in my mind, really is just Amazon — whether we're talking about what Amazon is doing in terms of giving customers near-unlimited choice across their own portfolio product and across the marketplace.

I think the latest count is over 350 million products that consumers can choose from, and we're talking about the expectations that they've said on convenience with Prime and one-day delivery. In many cases, in many urban areas, there’s even two- to three-hour delivery with Prime Now.

Even when we look at what we're doing with the world of logistics, they’re setting up their own fleet, setting up distribution centers at the back of the field under a tent. Really, what they're doing to drive disruption is incredible.

And it’s not just setting a new bar for what consumer expectations are across all channels, right? We're being measured against that bar of “What can Amazon do? What can Prime do? How quickly can I get that product? How conveniently can I get that product and how big is that assortment?”

But it’s also setting an entirely new bar in terms of what's feasible, what's possible in the world of digital-native companies like Amazon when it comes to delivering a truly unique and cost-effective experience across the supply chain.

So I always anchor on that one, and I think that's been — from what we've seen from our clients, and in my experience previously at McKinsey — hands-down the biggest disruption on supply chains, and the bar that folks set on themselves when it comes to delivering. Again, an effective and truly incredible experience with the supply chain.

Breen: I couldn't agree more, Fred. So much of the consumer expectations these days, so much of shopping behavior these days, and so much of the ways in which consumer goods manufacturers must respond to those changes in behavior really were driven by Amazon. If they didn't invent it themselves, they were the ones who really made it popular. So I think you're spot-on with your thought there.

I think that also provides a good jumping-off point for the rest of our discussion today. For decades, consumer goods companies were really good at marketing and selling their products.

The best of them also excelled at effectively planning how much of those products they were going to be able to sell, and manufacturing the right amount of those products. And then expertly moving mass quantities of those products to and through retail stores to be sold to consumers.

Supply chain expertise over the years became a critical aspect of profitability and also of growth for both manufacturers and retailers. And, in a lot of cases, it even became a very strong competitive advantage for some companies.

Omnichannel has forced companies to make some very critical adjustments. Adjustment is not even a good word because we're really talking about transformation here. Everything that Fred just outlined about Amazon all of the changes that have been taking place — endless assortments, ever-expanding arrays of options for fulfillment, extremely heightened expectations for delivery speed — forced many companies to really take a hard look and make a lot of significant changes to their traditional practices. And it's going to continue to be a critical aspect of success.

John Rossi, who was a consultant with SEA Path and a member of CGT's Executive Council, noted in a recent article that in the industry it's become pretty trendy lately to talk about improving customer experience for online shopping, and for creating a seamless omnichannel experience for consumers.

But what many companies have failed to realize is that they need to properly connect those efforts with the actual supply chain in order to ensure execution, because the vast majority of complaints that come from consumers are related to supply chain issues, not really experience issues. If you can't ship them the product they want when they want it, it doesn't matter how great your website customer interface is when they place that order.

And, so of course, these changes that we're talking about go all the way back to manufacturing because the efficiencies around mass production of flagship products must now be combined with customized items and smaller runs that reflect the changes that we're talking about here in consumer buying behavior. Companies are now finding and producing a small run of an innovative flavor and selling it directly online or maybe offering it only online is more efficient than seeking mass retail distribution at least in the beginning.

We're seeing this a lot, especially in terms of new product launches where companies are fully testing consumer response before making any big plans for physical distribution. And in some cases, brands are even finding it more efficient to manufacture purely on demand which we've seen in some areas of the apparel.

Whatever strategy a brand might be taking with manufacturing, success is going to depend on data. The information that's going to be used to fully understand demand and to truly optimize ordering and fulfillment.

And here's a quote from the report: "It really comes down to understanding our customer needs and enabling on-demand personalized accessible products, services, and experiences." that comes from Neil Akerman, who's head of advanced technologies for the global supply chain in the Middle East and Africa for Johnson & Johnson. He's also a client of Aera's.

Now I just wanted to point out the image here on the slide. This is an up-close shot of the fly print 3D printing technology that Nike is using to manufacture custom-ordered shoes that consumers are designing themselves online. So just a pretty strong illustration of the changes that are taking place.

So now, be that as it may, if this is the case of what's taking place in the industry, I would like to ask our experts at Aera to discuss it a little bit. With all of factors to consider in the manufacturing process now, what is the resulting impact on planning for a consumer goods company? Arnaud, if you could offer some thoughts here.

Arnaud Morvan: Yes, so as you mentioned, Peter, and as mentioned by Fred before, now we have things that market change led by the digital company. So it means supply chain will become more and more complex. We have larger scale, larger portfolio, and consumers want more personalization.

And he mentioned we need also more agility because you need to act more quickly to the demand changes of the consumer. You want even more speed.

So what it means that all these complexities, the scale, and this agility has impact how you plan and even replan for your manufacturing. Specifically, when you plan, it's a few weeks ahead, these static cycle times. Now CPG companies and retailers need to be able to replan their manufacturing in a matter of two days or even sometimes in a matter of few hours within this plan window. So it has synergy in platform to build to the work of the planners and to predict uncertainties.

Breen: The next area we wanted to cover is fulfillment, and we've already referenced some for the key issues going on here that kind of boil down to new channels for product purchase and methods of product fulfillment.

Another person we speak to in the report, Greg Buzak, the president of retail technology at analyst IHL, says, "Retail is very good at moving cases. It's not very good at moving eaches." And, of course, that term refers to e-commerce fulfillment, when brands are often shipping a single order even a single product to either a store for pickup or to the buyer's home or even elsewhere.

We've seen some experimentation recently where companies are letting consumers specify exactly where the delivery should take place: the office, the gym, the restaurant where the birthday party is taking place, the hotel where the consumer is going to be the next day, a lot of different options.

On the retail side, retailers are looking at their stores as distribution center, building dark stores solely for online fulfillment and constructing new warehouses that can manage both the traditional retail cases and these new single orders. Manufacturers are doing likewise and also grappling with the need to localize more of their operations in order to speed up delivery times.

Now to help address these issues, Gartner predicts that by the year 2023, more than one-third of operational warehouse workers are going to be working alongside robots. And that prediction leads us to the next discussion are: How will the need for automated processes, in general, and robots in particular, affect the traditional workforce? Fred, if you could take this one, please.

Fontes: I think you hit on a couple massive points there, Peter. In that, we're seeing organizations, honestly, not unlike the disruption that we're seeing in planning as a result of B2C and omnichannel fulfillment models. You have to completely rethink, not just their networks, but their facilities, their tools, all the way down to the processes of how they manage to deliver, your point, eaches all the way down to the consumer, where traditionally they delivered pallets to DCs or retail stores.

And in the case of omnichannel in particular, we're seeing not just the challenges of that single direction supply chain from the manufacturer all the way to the consumer. But you're also having to see the management of that end reverse. If you look at the returns phenomenon, especially in the apparel industry, I think last year alone was estimated to be a $50 billion problem over the holidays, much less the entire year.

So as a result, you're seeing exactly the trends that you discussed on the network side. Companies are looking to completely redesign their network and think about it in really innovative ways. You mentioned dark stores and fulfillment centers that are really managing both that direct consumer and the store piece. And within those changes in the supply chain model or the network model, seeing a lot of innovation when it comes to the use of automation technologies.

What I think is inspiring as we look as some of the technologies that are being developed is that they're not really focused on: “How do I replace the human labor in those particular DCs or warehouses or across the fulfillment network?” But because we're going down to, I think, a space where robotics still hasn't caught up to human dexterity and human cognition, the need for those robots to be able to collaborate with a human, right? Because we're going down to eaches and we're going down to parcels that are not necessarily all of the standard shape and size that are easy to recognize.

I think you also see on the flip side that companies like Amazon are investing massive amounts of money into what they call the softer side of robotics. And we'll some advances in those space, but at least today, I think we're seeing a lot more emphasis on that human-machine collaboration and interaction. And a lot of emphasis beside that on the upscaling and rescaling of the human labor to be able to control and to manage those robots. And let the robots do what they're fantastic at and the humans control some of that higher-level cognitive function.

Breen: Could you maybe talk a little bit about some of those higher-level functions specifically that the humans will be able to focus more on?

Fontes: I think we're seeing a lot more focus on the types of activities around recognition of, again, these eaches and parcels that are maybe a little bit tougher for computer vision or for robotics to do today. You're seeing a lot more focus on the humans guiding the decision-making of the robots based on more information. And being able to also, I think, tribute back a little bit more to the intelligence of the robots that machines are using.

Not to steal too much thunder from the next question we're going to on last mile, but there's a really interesting study from MIT recently where they've really started working on this collaboration within human and machine where the machine is developing, let's say, an optimal route to fulfill within a city. And the human is providing feedback: “Yeah, actually, that route might be optimal, but that particular street that you're telling me to turn into is usually incredibly crowded or is difficult to park on.”

Again, that human-machine interaction and learn from the best of both worlds that optimize algorithm that it can be the best route that I can possible take combined with the human knowledge that, again, is inherently difficult for machine to gather just based on data or inputs lie computer vision.

Breen: That's an interesting example you use because I feel like I have that battle with my GPS almost daily. It's telling me to take a different route, and I'm trying to decide whether or not that's the best way to go. It’s always a struggle getting across the George Washington Bridge for that reason.

The next area is the piece of fulfillment that is commonly now referred to as the last mile. And as I was thinking about it, it might more accurately be called the last new mile because, for a lot of product manufacturers, these are fulfillment requirements that they never really had to worry about before. And again, mostly things that we've mentioned already, but on-demand ordering from retailers and drop shipping to consumer homes or anywhere else they might be asking for the delivery to take place.

Here’s a quote from the report from John Harding who is the global CEO for Conair Corporation: "Our vision of last-mile delivery is ultimately to provide our products to consumers when and where they want to receive them regardless of where the ordering takes place whether it's a retail partner site or a direct-to-consumer site."

He says, "We aim to provide the best possible customer experience where initiated by a drop-ship out of our omnichannel fulfillment centers."

We'll talk a little bit more about what he means there in a few slides, but I think that just generally sums up the goal for a lot of companies right now. As if building out the infrastructure for this kinds of capabilities isn't challenging enough, manufacturers also then have to change their demand planning practices in order to address these changes.

So I'd like to drill down a little bit into the question that we just asked and ask our experts if they were to discuss a little more pointedly, what is the impact on planning here? How do you plan for eaches?

Morvan: I think it takes us back to what we produce in the case about manufacturing. Delivery of an item’s complexity, I think, is going to increase because now you have smaller and smaller packages to deliver to millions of consumer. You see Amazon trucks everywhere. It requires significant changes of how you can address the last mile. So it's, therefore, an omnichannel world. We tell CPG companies that they need more and more proximity with the consumer.

An example is one we gave before about the doctor. And basically, now you have this network of these small fulfillment center. So the first consequence is the increase of the complexity of your supply chain. The second consequence you have is that you need to extend visibility of your supply chain down to the consumer level. You need to know, for example, when your package is delivered to your home.

The last one, the impact on the planning, is in automation because the volume of that is so large to process, especially for last mile, that you cannot do it with a human being. So you need to automate this planning of the last mile as well, and therefore the specific algorithm — the one mentioned by Fred before for last-mile delivery — that takes into account input from humans.

Breen: The last area to cover, and we've already been discussing this, but to drill down a little more because it is so critical to future success: Arnaud, two phrases that you used in your comments I think lead us into the discussion here. You mentioned greater complexity within the supply chain. And you also mentioned an increase in data volume to help make sense of that complexity. So if we could talk about that for a little bit.

Traditionally consumer goods companies leaned largely, and in some cases almost exclusively, on a product sales history to drive decision-making. It was reactionary. It examined sales data and shipments and orders and market data. But today there are technologies like artificial intelligence that are not only getting manufacturers better insights into shopping behavior because they could look backward but also forward. And they can take into account a vast array of internal and external factors that go well beyond the historical sales data. And this, of course, as we've been saying is all taking place in an environment where time has really become of the essence.

In just one example from the report, Conair says that it's now looking at real-time availability at inventory levels as well as social media analytics to predict buying trends. And then it layers that kind of information in with the more traditional retailer point of sale data and market research and sales history to build out smarter demand planning practices.

Those kind of practices have become possible because of the new tools and technologies that are available. That's where I wanted to touch on here and wanted to ask our experts at Aera to talk a little bit about the impact that artificial intelligence and machine learning is having on the supply chain and the ways in which manufacturers are transforming their traditional practices.

Fontes: I think again, you've hit the nail on the head in the report and the scalability and the availability of compute that we have as a result of today's cloud providers. Plus, I think the advancement, excuse me, that we've seen in the types of algorithms that are deployed against this forecasting problem has really given us, I think, more interesting capabilities than we've ever had before.

We can be much more granular than we ever were before with our forecasting. We can be much more frequent than we ever were before with our forecasting. I think it's not uncommon to hear about companies that are now refreshing their forecast not just on daily basis, but even grains below that because they can see these changes in demand from, like you said, indicators, from media feeds and other outside factors.

But also, again, to that point, we can incorporate many more causal factors and external data than we ever did before. Part of it is the availability of data that we have. I think another part of it is, again, the sophistication of the algorithms have been developed that can adjust that data. They can make sense of it and come up with a new forecast.

What I think has been really interesting again in this space, in particular, is this idea of human-machine interaction. How do we make the absolute best of that new forecasting algorithm that's taking into account many more factors than a human ever could and come up with a forecast that is more objective in many cases than folks might say their sales team is talking about or their demand planners are thinking about as they're planning their business.

But at the same time, it incorporates unknowns or external data factors that maybe aren't coming in through a structured data feed or through social media but your human planners are aware of. And as a result, again, it combines into something that is better than the sum of its parts if you will.

That said, I think the flip side of all of these advancements and forecasting — in particular, the use of AI and machine learning — is that is any planner or forecaster in the business knows a forecast is a forecast and will always be wrong for some matter of degree.

What I think we're finding is much more interesting in this space. It's not just how are we applying advanced analytics, AI, machine learning, name your buzzword du jour, to really derive the forecasting process. But how are we also applying these technologies, scalable compute, and new ways of thinking about the process to react much more quickly when we find a forecast is absolutely wrong?

I can tell you a couple of days into the week if we're going to meet our targets for that particular week. Now I've changed my forecast. I've changed my expectations. I don't need to necessarily wait for the next S&OP cycle or planning discussion to respond and react when I've got, again, technology at my disposal that can really help me react, respond and drive the change that I need to catch up to that change.

Breen: In our next section we'll look at a couple of the solutions that consumer companies are implementing in order to help this transformation along. Gartner expects that by 2023, 50% of all large global companies will be using AI, advanced analytics, and internet-of-things tools within their supply chain operations, just to name some of the tools that are being implemented.

We already heard about what Gartner said about robotics, and we discussed that a little bit. But one key area that's a possible solution for companies that is very near and dear to the hearts of the people at Aera is cognitive automation. And I'm going to turn things over to Arnaud to talk a little bit about that concept and what it means for consumer goods companies.

Morvan: Cognitive automation is a very new term, and I'm sure most of the attendees of this event have probably never heard of it. But we have talked about it for two years now, and in many ways it's actually the reason why Aera exists.

Let me start by the definition. Cognitive automation is the automation and augmentation of decision-making and its execution and complex business processes. So what it means to this is the idea of moving from people doing the work, like manual work in Excel, as it is by to machines doing the work guided by people.

So the definition of cognitive automation is to have better visibility of operations, then with maximum of derived process to augment and automate decision-making. So it's basically what today is being done by multiple planners doing 50% of their time when they go in Excel to process and message decisions. And ultimately with cognitive automation, you can monitor the execution of the decision. So, over time, you can improve decisions that are made. And for that, we use machine learning and AI.

Breen: The next topic, and one that's covered a little bit in the report is RFID, which is driving greater accuracy within the supply chain for the companies that are already using it.

And the example side in the report is a partnership between GS1 and a reticent that study the impact of RFID technologies on a smaller parallel company called Southern Fried Cotton. This company had been experiencing a lot of chargebacks from retailers due to inaccurate orders, but by implementing RFID to track the products through the system more accurately, it was able to almost eliminate the issue entirely and now claims to have 99.5% accuracy on its orders.

It should be noted that this company was moving from a really old-fashioned paper-based recording system. Most companies these days are probably a little more advanced than that, but it still does illustrate the level of improvement that you can make through the proper use of technology. Fred, can you to talk a little bit about RFID and its potential impact?

Fontes: I think we're seeing our ideas and other major ideas unlocked, as you said, not just to the ability across the supply chain — which has become increasingly critical as we move, again, from a world in which I got a handful of warehouses and a couple hundred items. And I can relatively easily design a process by which I can check my inventor and I know where it is.

It’s moved to a world in which I've got millions of product and I'm delivering those product, not to warehouses, but to millions of consumer's homes. You need some sort of technology that is able to track where those items are, how they move through the nodes accurately and precisely.

Those processes that we design for the warehouse and the old-school networks are still in scale. And we're, again, seeing our ideas of massive improvement. To be able to drive that visibility, you're seeing as a result of the reduction in cost.

And RFID technology is being implemented in industries where it wouldn't have been feasible just a couple of years ago. Whether you're talking about supply chain, IT investing heavily in what they're doing with RFID as I think from a visibility standpoint, it's been a massive unlock for suppliers to understand where their product is across the nodes.

But then, on the flip side, also I think it’s a massive improvement in terms of the efficiency by which we can measure where the inventory lives. And with RFID technology I no longer need to rely on those heavily manual processes, weekly inventory checks, if I know exactly where an item is and how it has moved through my network. And, again, that being just so critical today when we're moving from a handful of nodes to thousands or millions of nodes if we think about, again, that delivery to the consumer and that last-mile component.

Breen: Is it really the decrease in cost that has RFID being used to the extent it is now? It's not a new technology, right? It's been around for a long time. So the potential was always there, but has the industry been waiting for the cost to drop to the level where it is feasible?

Fontes: I think you've seen on one side it's a combination of the cost dropping to where it's economically feasible and, again, justifiable for the problem that it's trying to solve. Again, the visibility across the network and the efficiency of gaining that visibility. I think the flip side too is it's a more ubiquitous technology when it comes to integration with upstream systems now. I think a few years back you would have been hard-pressed to find solutions across supply chain management that can natively integrate that data coming from the RFID into your end-to-end visibility or planning solutions. Where nowadays those integrations have become a little bit more seamless.

Breen: Our final solution is data. And we've been talking about this a lot, but the idea of more and better data to drive better decision making. More sources and streams to help companies better understand consumer behavior and then better tools to synthesize and analyze the massive volume of data that's now available to manufacturers. As we heard, Neil Ackerman from J&J say earlier, "Success really is all about the data these days."

Arnaud, if I could ask you to offer some thoughts in this area.

Morvan: You just mentioned RFID, but you also have ERP, CRM, IoT. So clearly there is no lack in the proliferation of data sources that's available today. But that said, when we go to see a lot of companies that really struggle to develop the value of this data. And even better that sits within the folds of the company, and we can still have lots of companies that whether it is a silo difficult to access. And, as a result, they are underused and they don't bring value.

I think the key now for lots of companies is to make all of that happen and to get as much value as they can. I think one of the keys now is how to process all the data and to make them insightful.

Fontes: I think we're also landing in a bit of a world, Peter, where consumers are starting to get a little bit more wary about the data that's out there on them. I think we've had a sequence of articles whether it's in The New York Times talking about GPS data aggregation, or where I believe it's Fox talking about a company that was supposed to be protecting your data, your anti-virus actually selling that information off to third parties. We've talked a lot about this proliferation of data and where we can go with all of these exotic new data sources that the world is making available to us — whether or not, again, their uses may be crossing some boundaries that we need to consider as a group of proponents for more data.

But, to Arnaud's point, I think we continue to find over that the majority of the value to be unlocked for many of the organizations we work with — sites with data that's within the four walls of the organization and has been either underused or not readily available for their users and their processes and their analysis — in many cases just getting to that internal data happens to be a much bigger driver of value than looking for that data source that is going to add some marginal value to how we think about the consumer and, again, maybe crossing some ethical lines.

Breen: So unlocking what's already there before you even go looking for what else might be out there. That's very interesting point.

Fontes: Exactly.

Breen: We wanted to spend a little of time next on a couple of case studies. And the first one comes from the report. We've already heard a little bit about what Conair Corporation has been doing. But the transformation that Conair has been undertaking cuts across all of the topics that we've been covering here today.

In terms of demand planning, Conair has been working to build out capabilities that let them respond to all of the new marketplace requirements; balancing inventory to address the traditional needs of pallets for the brick and mortar store, and being able to fulfill single order eaches for online orders.

We already mentioned the newer data sources that Conair is using looking at real-time availability of inventory levels, along with social media analytics to predict buying trends and then combining the new data with the old data and the POS in historical sales data to build a smarter demand planning system. In terms of fulfillment, the changes that they're undertaking in Conair involve making a pretty significant investment in a new omnichannel distribution center.

Two years ago Conair opened an 800,000-square-foot DC in Glendale, AZ, specifically to manage smaller e-commerce orders. The facility is adjacent to two traditional centers that still handle the old fashion pallet level orders.

And within the facility among the more cutting-edge tools that they're using are voice-activated picking technology instead of the traditional handheld scanners. Building out these DTC capabilities for e-commerce in conjunction with their retailers has let Conair start its own direct-to-consumer e-commerce business, which right now it's been using largely the focus on products that are less likely to be available through traditional stores like spare parts for Conair's kitchen appliances.

But it's also leveraging these capabilities as an initial launch platform for some of its innovative new products that it might want to test a little more before rolling out to physical stores.

One final thought from Conair CIO John Harding. It might sound a little simplistic, but it really, again, goes to the heart of everything we've talking about today: "The goal with all the change that's taking place in the company is to provide our products to consumers when and where they want to receive them." Sounds pretty simple as a concept, but of course, we know how complex it is to actually undertake.

And now for our next case study, I'm going to turn things over to Aera to talk about a CPG company that is implementing touchless forecasting and supply response.

Fontes: This case is a little bit different. We're going upstream to talk about a company in CPG, again, in this case, a $10 billion company that we've been working with for quite some time. That's doing really interesting the things when it comes to leveraging cognitive automation, the concept that Arnaud was talking about earlier to really move from humans doing the work a little bit of assistance from machines or tools to really machines doing the work with guidance or input from people.

So the background here is many companies are going through today — and I'm sure many of our audience members are going through — this organization was embarking on a massive, call it, digital or advanced analytics transformation across their supply chain and looking to really up-level what they were doing in terms of the forecasting. But also how that forecast was impacting not just their long-range planning, but really the call it short-range firefighting if you will that their teams were doing on a regular basis.

Some of the biggest challenges that the client was facing when trying to really tackle these problems were, again, as we see in many organizations we work with an eco-system, underlying systems that were disconnected, disparate, not necessarily updated to the latest version. And as a result, made it very difficult to gain end-to-end visibility across the business.

Pressure from their external suppliers and customers to be much more agile and responsive, I think the quote that always sticks in my mind is, we went from interacting with a buyer at our retailers and we had predictability on how they purchased to an algorithm telling us what they needed were Amazon supply, triggering orders and patterns that we'd never seen before. And as a result, had a hard time managing.

And then, lastly, IP organization that was swamped with dozens of other projects and other focus areas and made it difficult to really drive this analytics transformation. So what we helped this particular client with as Peter mentioned is two topics. One, touchless forecasting, and what we've been doing there is helping them ingest and harmonize internal and external data across systems to paint that end-to-end picture of demand and all of its drivers whether we're talking about internal data on order history or we're talking about external data coming from third parties like a Neil center or an IRI channel inventory coming from the retailers that they work with. And being able to, again, ingest, process, harmonize all that data. And then feed it into a series of automated algorithms that are projecting, predicting, forecasting what the demand is going to look like.

But not only doing that but really evaluating what's the accuracy of the algorithm? Which algorithm performs best for a particular product and location going down to a much more granular level then they had before? And then being able to pick, again, that algorithm that works best and be able to automatically publish that forecast all the way down into their downstream systems. The result of this is that over 50% of their portfolio now gets forecasted to what they call completely touchless. Meaning that absolutely no human is involved in the process all the way from data collection, cleansing, preprocessing, forecast development, all the way down to the publishing of this forecast into the downstream systems.

The other 50% is, I think, the more interesting part which is that human-machine interaction which is where we're taking products that are more difficult to forecast, maybe are more volatile than their core, stable products and we're trying to take the best of both worlds. How do we take insights from the algorithm and marry those with insights from the planners to come up with a better forecast at the end of the day?

In all of these, what we're doing in the backend though is we're continuously tracking what is the performance of the underlying forecasting algorithm? What's the performance of the planner changes and overrides? And using that to guide the forecasting process going forward.

The second piece though that I think, again, is the much more powerful portion for all of those who've been involved in these kind of forecasting or planning processes in the past is what happens when the forecast is inherently going to be wrong. And that's where we've been working with the client on what we call supplier response or cognitive supplier response where the machine, in this case, is looking at every single imbalance across the supply chain all the way down at the product node, product, think about it as ship-to location level.

And after looking at those imbalances, there's a value in every single possible option to resolve that imbalance whether it's transfers of supply between locations, whether we're talking about increasing production capacity all the way downstream at the plant, whether it be expediting product that's already in transit but needs to move from one mode to another or get prioritized.

And essentially doing what the planners were attempting to in the past in their firefighting but being able to do it at scale in an automated fashion completely unbiased of the situation, truly following just what is absolutely best for the business from a margin and an impact standpoint. And being able to cover every single product under that portfolio, not having to apply 80/20 rules to really focus on the things that are highest priority because that's the only thing that time in the day permits for.

So this has been a super exciting journey that we've been on with this particular company and excited to see where we can take cognitive automation next.

Breen: Did you meet any resistance with the internal stakeholders or was the company really ready to go and embrace this technology?

Fontes: I don't think I would call it resistance. Probably the most interesting challenge or question we've had to deal with in the process is now the one of who owns the forecast and if the IT that is managing the platform that generates the forecast? Is it the data science team that is adjusting the algorithms? Or is it the planning team that occasionally overrides what the forecasting algorithm is providing?

As you can imagine at a CPG company, where the forecast drives massive amount of activities that have all kinds of upstream and downstream impacts, it’s not a question that is easily answered. But I think we're hitting some interesting territory and some interesting ground in terms of what does it mean to own the results of an automated process and AI-driven process and how do we think about organizing differently under that construct where automation is going to be of the norm than it is the exception?

Breen: We're heading into the home stretch here, so I just wanted to talk a minute about the future. I don't know if this picture accurately represents the future or not. I'm, I think, hoping that it isn't. But what you see here is the driverless mobile store that Stop & Shop has been testing up in Boston near the concept here.

Shoppers request a visit through an app and then the store comes to them with whatever products the retailer feels would be best suited to the driving around visiting people. Not quite sure what kind of wrinkles this would add to the existing supply chain, but so many of these kinds of things out there being tested now. So it will be interesting to see where the industry finally heads.

But on that note, I wanted to circle back to the first question that we asked upfront about the most significant disruption that we've seen, and ask our friends at Aera to consider. What will be the single greatest solution to the disruption that's taking place in the traditional supply chain over the next five years? And Fred, if you can take that.

Fontes: Yeah, absolutely, right after I tell you that forecasts are mostly wrong, you asked me to make a forecast. I give my best stab at predicting the future on this one. I think from our perspective if you look at what has happened in supply chain tooling and over the past few decades when it comes to automation, we've gone all the way from automating the manufacturing floor that has moved back to automating many parts of fulfillment.

We're looking at automation in many cases all the way through to the actual detail store or the robot that's fulfilling the order in the last mile. And we think really that trend of cognitive automation is the next horizon of that. By taking automation out of the delivery or execution activities and moving the automation up to that cognitive level, the decision-making level within the supply chain.

And what we're really excited about in that space is the kind of opportunity that it's going to unlock for the companies that implement these changes by being able to be much more granular under decision-making to improve the accuracy of their decisions to move much more frequently in an automated fashion across their business and to do so again with end-to-end coverage. And again, the opportunities of that I see much more responsive to customer changes, being much more responsive to trends that are happening in the industry, being able to hyper-personalize the product that they make and deliver to their customers. These are all, I think, opportunities that these companies are going to be able to unlock and explore. But it's going to require a degree of automation of decision-making that we haven't seen before.

Breen: That was great answer to the tough question. We do have a couple of minutes for some questions from the audience now. So we'll jump right into that and if we can start off. One was related to your case study. The question is, what are the technical requirements and resources that would be needed to implement what you were talking about there? And specifically a question around data scientist headcount and use of RFID. So Arnaud, could you maybe take that?

Morvan: Yes, I completely agree. First, you have new roles in companies that you have to approve for. The key is all up to people who can understand it. So it's very new type of profile that’s pretty difficult to find. It depends on how far you want to go with your team. If you want just your basic run-better sales, you just need two people. If now you want to have a lot of science in how you run your business, of course, you need more.

One of our customers … is a team of around 30 data scientists, but it's across multiple project around for the global supply chain. And on RFID, you have two impacts. The first one is not internal; it's with your supplier because now you need to offer the supplier to be sure they put the RFID. And after internal, it's how to run your operations, especially your DC operations with RFID as well as the store. So you have both impact internally and externally on RFID.

Fontes: Yeah, I think, the piece that's really interesting in that space is on one side, Arnaud is spot on. You're having to think about completely new roles that I think especially in the realm of CPG are maybe foreign to some of the organizations that are trying to tackle these new kinds of opportunities around AI, machine learning, cognitive automation.

I think on the other hand it's also the question of how do you get those resources to truly be productive and to be able to scale their thinking across the organization. And that's where the alignment between business and IT is incredibly critical. I think where some organizations fall short in that journey is thinking that because they've hired a cadre of data scientists and engineers and UX designers and business translators, all of a sudden magic is going to happen.

I think you need quite a bit of alignment between business on the use cases that you need to go tackle and IT on how they're going to support and scale the work that organization is going to do in order to really be effective.

Breen: Beyond RFID, are there other types of technologies that you see gaining steam to help out in these areas?

Fontes: When we're going from RFID being applied to many more products not used because cost is going down, the question of will it be or can it be applied to eaches or lower-cost items when we're talking about growth? And what are some of the new technologies that we're seeing?

I think you're seeing a lot more focus on computer vision for those types of problems because it's a more scalable solution, right? So whether you're talking about the Amazon Go stores or all kinds of companies that across the supply chain are looking to automate, you're focusing much more on that computer vision.

I think a great example is Ocado as they're thinking about their backend supply chain, how they manage delivery of grocery type items at scale. They're finding that RFID works all the way up to the point where you have to go from the case to the each.

And then from the each of that manages forward across the supply chain they're using computer vision to be able to track upon these items that move through the different nodes. Because while we love the fact that RFID has certainly gotten a lot cheaper over the last decade, they're still not down to the point where on a dollar or below item they're going to be cost-effective to manage when they're already treading on razor-thin margins.