Christina Cardoza: Hello and welcome to “insight.tech Talk,” formerly known as “IoT Chat,” but with the same high-quality conversations around IoT, technology trends, and the latest innovations you’ve come to know. I’m your host, Christina Cardoza, Editorial Director of insight.tech, and today I’m joined by Matt Redwood, Vice President of Retail Technology at Diebold Nixdorf. Hey, Matt, thanks for joining us.
Matt Redwood: Hi, Christina. It is great to be here speaking with you today.
Christina Cardoza: So, for those of our listeners who are not familiar with Diebold, what can you tell us about the company and what you do there?
Matt Redwood: Diebold Nixdorf is a technology company of two halves. We provide banking systems to the world’s largest banks, and we provide retail technology to the world’s largest retailers. I’m responsible for retail technology. We provide hardware, software, and services to most of the 25 top retailers globally, as well as quite a few tier-two, tier-three retailers.
And we generally cover front-end technology, which we’ll go into more detail on, software and enterprise software. And then we provide most of the services—break/fix and help desk services—to retailers to make sure all their technology is up and running for the maximum time possible.
Christina Cardoza: Great. And obviously we’ll be focusing on the retail aspect of Diebold Nixdorf today. We’ll have to get someone else on a later podcast to talk about the financial aspects of the company.
But the last time we spoke with you, Matt, it was for an article on insight.tech, and we spoke about POSs transforming retail checkouts to improve customer experiences in stores. But customer experiences—I think that’s just one pain point that retailers are facing today, one challenge. So, that’s where I wanted to start the conversation off today. What are the different challenges retailers face today, in addition to customer service, in stores?
Matt Redwood: So, it’s a bit of a tough time for retailers. And I think, regardless of what sub-vertical of retailer you are in, I think most retailers are struggling with the same challenges. So, on one side, as you said, customer experience is key—the desire or drive to make sure that the in-store experience is as high as possible with this ever-changing horizon or landscape amongst consumers, that their expectations continue to rise. So, the horizon of expectation continues, and retailers are really chasing after that. And we are really starting, post-COVID, to see retailers really investing again very heavily in that in-store experience, which is great to see.
On the flip side, on the top line and bottom line, they’re being squeezed. I think you can read in the press global and economic trends that are driving the cost of goods up, the cost of freight up, the cost of managing and running their stores up. So, their top line is being squeezed, their bottom line is being squeezed, and they must find ways of driving efficiencies in the store while also delivering that great consumer experience. It’s a real balance between getting the economics of retail right, as well as satisfying the needs of your consumers.
And competition is as high as it’s probably ever been in retail, which is good in certain aspects. It helps with pricing and keeping inflation under control. But on the flip side it means that consumers are very flippant in terms of where they get their experience from and where they shop. If they get a bad experience in a store, it’s easy for them to flip to another brand and get a better experience of potentially better products, better prices. So, it’s a very dynamically changing environment, very difficult one for retailers today.
Christina Cardoza: I’ve seen a lot of retailers start adding new technology, more intelligent technology and sensors, to be able to do some of these things: collect data at the edge in real time so they can make decisions as they’re happening. A lot of this is being powered by artificial intelligence, and I think we’re in a stage or a point today in the industry where AI is everywhere, and everybody’s trying to use it and get the benefits from it.
So, from your perspective, how is AI being able to address some of those challenges that you talked about, and what’s the reality of it? What are the real benefits that are coming? Because I feel like sometimes there’s hype, but where can we start using and getting actionable insights?
Matt Redwood: I think 2023, for most people, will be known as the year of AI. It’s where generative AI really took off in retail, and we started to see more and more AI applications in the retail market. And certainly, some companies really jumped to what I would consider the end goal of AI—which is completely changing the technology landscape, completely changing the customer journeys, the staff journeys, how you operate and run your stores—with this kind of euphoric view that AI could remove all technology that existed within stores.
That’s what I call the hype curve. We’re coming through the trough and we’re going back up again, in that a lot of people realize that that technology, although fantastically advanced, was probably quite a way off being realistically deployable en masse. The cost of the technology was high, there were limitations in terms of the size of the store and the amount of products and the amount of consumers. So, trying to take that technology and apply it to retailers today wasn’t applicable.
So, what we are seeing, and what a lot of retailers have done, is kind of take stock of the situation, re-address what’s really important, focus in on the pain points, and then really go, again, with what we call point-solution AI technology: so, specific AI deployed for a specific use case to solve a specific problem, but is very much for that particular use case. And we’re starting to see more and more of these solutions being trialed across retail stores, not only in grocery.
And the possibilities are really—they are bountiful, and they’re kind of endless. And some of the examples that we’re seeing are everything from health and safety in store—using AI on top of CCTV networks to make sure fire exits aren’t blocked or there’s not foreign objects or liquid spill on the floor where someone might slip over. We’re using it for heat mapping to understand what is the flow of consumers around stores—how do I make that flow easier, but also how do I potentially commercialize that flow?
We’re seeing AI on top of existing technology—so, something very close to my heart: self-service. We’re starting to see more and more AI being applied on top of existing technologies to make them more efficient, to make them easier to use, to close loopholes, to boost the consumer experience. So, technologies like facial recognition for age verification.
I think we’ve all been in the situation where we’re trying to buy paracetamol or a bottle of wine, and you must wait for a member of staff to come over and approve your ID. That’s been compounded, the effect of that situation, by the fact that retailers are struggling to find staff. So now I’m having to wait a little bit longer to have a member of staff be available to come and approve my ID. Using AI in that environment drives greater efficiency at the frontend. It reduces that requirement on members of staff, and it boosts that consumer experience.
We’re also seeing technologies centered around the product. Item recognition, really, really taking off. Not just for non-barcoded items—where we’ve seen fruits and vegetables selection—but also all item recognition. And in some environments, particularly smaller stores, why should you have to scan the barcode when you can identify the item by its image? So that’s an exciting technology.
And then finally, something that we’ve been working on over the last 18 months, which is anti-shrink technology using AI. Obviously shrink is something that’s really gone through the roof in a lot of retail environments, driven by the cost-of-living crisis. And we are now working with a lot of retailers; we’ve got 54 different retailers we’re working with on anti-shrink technology in one form or another to try and close those loopholes and make it more difficult for those that are maliciously trying to steal, making it difficult for them to be able to steal. But also those that may have just been unfamiliar with a process or genuinely have made a mistake. Also making sure that we are catching that, without making that a bad experience for that particular consumer.
Christina Cardoza: It’s interesting; in the beginning of your response you mentioned how retailers, they were adding this technology to really transform everything, and they were sort of jumping to the end. And especially when you’re implementing artificial intelligence, which has so many connotations with it, so many misconceptions. It’s interesting, because I feel like these things need to be gradually introduced to consumers for them to be able to accept it, to understand it, to use it.
I can’t tell you how many times I’ve been in self-checkout, where we’re using AI or computer vision, and I can’t even put an item on the scale after I’m done scanning it because it needs to be in a bag, or I can’t bag it yet because of the bag weight. It’s just so complicated.
I know every retailer has different challenges and different areas of entry, but would you say there is an easier place of adoption happening right now to adding some of this intelligent technology? And then not only easier to adoption for consumers and for the store, but—like you were talking about the facial recognition—I know consumers have privacy concerns around that. So how can stores easily implement this that makes the most sense for consumers and for themselves and their business?
Matt Redwood: Sure. So, complex question. I’m going to break it down into parts. So, when we talked about retailers and some technologists rushing to that endgame, it really was about trying to boil the ocean with AI to try and completely change the landscape of retail. And I think sometimes what we forget is although the technology may exist, forget whether it’s commercially viable or practical to deploy it, you have to also have consumer adoption. If you don’t have consumer adoption, no one will use the technology in it. It’s worthless.
So we very much, we track the consumer-adoption curve, and we track the technology-development curve, and it’s important to find something broadly in the middle of those two in terms of what’s the right technology; what’s the right innovation and technology; why am I deploying it? Making sure consumers adopt it, but, crucially, making sure that it solves a need and it solves a business or a consumer desire. The “build it and they will come” mentality does not work with innovation, and it doesn’t work broadly with retail technology.
Consumers are savvy, and retailers are much, much more savvy in terms of deploying the technology. It must deliver. So we always recommend starting with data. A lot of people talk about data; there’s a lot of data that exists; it’s very easy to be swamped by data. We call it paralysis by analysis. There’s too much data out there. But if you can really segment your data to understand—if I’m looking at my transactional process or my customer journey—making sure that I’m looking only at the data that relates to that and highlighting the problems.
I’d say 98%, 99% of our customers that we work with now, we actually work well on a consultative basis to actually really deeply understand their stores, how they’re being run, and how their consumers shop in their stores. And the data provides a lot of insights to that. So really understanding and analyzing: How is the store operating today? Where is the friction associated with the staff journey or the consumer journey? Understanding and quantifying the effect that that friction really then builds the picture to say, “Okay, I’ve got a problem statement I want to try and solve. It’s having this impact on consumers and staff, and this is the impact to my business.” And that’s relatively easy to calculate.
The more problematic piece is then really finding the right innovation to solve that. And very much we try and put the consumer and the staff journey at the center of everything that we do. If it doesn’t provide value for the consumer, if it doesn’t provide value for the members of staff, and it doesn’t provide value for the retailer—that triangle of value is at the center of everything that we do. And if it’s not ticking all three of those boxes, we don’t put it into the range and we don’t put it into the solutions or the stores.
Starting with that data is a bit like the treasure map. It highlights where your biggest areas of inefficiency are and then provides the compass to kind of point you in the right direction of what’s the right technology that you should be deploying to the store to actually try and solve that particular issue. And when you break it down like that, and we start thinking about this kind of AI-boil-the-ocean vision, we start thinking about individual point solution, it becomes much easier because it’s much more manageable to deploy from a technology perspective, it’s much easier to develop a solution that works for a particular use case or problem that you’re trying to solve.
But it’s also then arguably very easy to measure how successful it’s been once you put it into the store. The difficulty then comes is what you don’t want to do is collect a whole group of point solutions that don’t talk to each other, and it becomes very, very difficult to scale. Finding the right AI platform that allows you to scale all of these point solutions on a singular platform is really, really important.
Christina Cardoza: Yeah. I love one thing that you said, which was basically, if it’s not solving a problem or if it’s not benefiting the customer or the business, then don’t do that. I feel like that’s a major problem that we have with implementing technology and seeing shiny new things. Let’s just add it to add it, but why are we adding it? It’s not going to get you a return on investment, and it’s not going to help your business if it’s not really doing anything for you. So, I think that that was a great point.
I want to come back to also that facial recognition example again—how obviously I think we’ve all dealt with self-service checkouts or checkouts where you’re scanning something, it doesn’t recognize it, you need a human cashier to come and help you, and that just bottlenecks the entire process.
But there seems to be a lot more self-checkouts in the store. How does the role of the employees come into this? I know, talking about the consumer misconceptions that they have, there’s a lot of misconceptions that this is going to replace employee jobs, and especially when you see that it—there’s not a lot of cashiers on the floor anymore. So where does the human element come into play with some of these?
Matt Redwood: So, the human element is really, really important to self-service, and it’s an element that’s quite often overlooked. If you look at the evolution of self-service, self-service was originally designed as a POS, an attendant till replacement, to ultimately remove the cost of the staffing from stores. But self-service has been around for 20, 25 years now, and the drivers for deploying self-service are very different today compared to 20, 25 years ago.
I’d say 100% of the retailers that we deal with are either putting in self-service—and they might be on their second or third iteration of self-service because they’ve been in that business for a while—or they’re putting self-service in for the first time. A lot of retailers outside of grocery are just trying self-service for the first time. The approach is very, very different, and it’s very much less about removing staff from the equation, more about staff redistribution. The inability to attract and retain staff in retail is a real big problem for retailers, so they have to use their staff wisely. And where the consumers value the staff interaction the most is where they need it, and where they need it is where they generally they need help either navigating the store, finding an item, asking a question about a particular item, or just general assistance.
What self-service is really playing a major role in retail today is it unlocks that member of staff. So I would say to anyone that looks at self-service and says, “Oh that’s going to replace people’s jobs,” it’s not; it’s very much about labor redistribution now. It frees up a cashier that could be sat behind a till for a 12-hour shift to be up on their feet, engaging with consumers shoulder to shoulder in the aisle where it really makes sense to deliver that consumer experience.
Particularly through Covid we saw retailers that that had self-service had much greater flexibility of operations within their store. Post-Covid we’re now seeing that that allows them to boost the level of consumer experience where it really counts. Obviously, there’s always been friction associated with self-service and the adage of “unexpected item in the bagging area”—all of those common friction points perceived with self-service, they’re starting to really drain away.
A lot of focus has been put on fine-tuning and making sure that the base technology works to a much, much more acceptable level. And we’re now seeing self-service that’s very efficient, that generally most of the time you can sell-through a transaction with no intervention, no requirement for a member of staff to come over. We are now in the fine-tuning era of self-service, and why I say fine-tuning is we’re really looking for that last 5% or 10% of efficiency gains.
So, Diebold Nixdorf, we’ve really focused on three core solutions initially out the bag, and those three core technologies have been developed because we identified via the data where the biggest friction points were. So, age verification: 22% of interventions broadly are age related. That’s a big number. If we can use facial recognition to identify the age of the consumer and remove that validation process that’s happening—A, much better experience for the consumer; B, it means faster transactions. Faster transaction means less staff requirement at the till, but it also means that consumers are moving through the frontend quicker.
So that means less queues. Less queues—queuing is the biggest bugbear of consumers when they get to checkout. So, we’ve removed two of the biggest friction points, associated with checkout with one piece of technology. Item recognition, particularly in grocery for fresh fruit and vegetables, was another area of frustration from a consumer perspective. But also inefficiency from a retailer perspective: spending 20, 30, 40, 50 seconds, trying to find the type of apples that I’m looking to buy is frustrating, but it’s also time consuming. So, using item recognition to identify those apples so the consumer doesn’t have to run that process again. Good consumer experience, great productivity gains.
And then finally shrink. We touched on it a little bit earlier, but obviously retail shrink has really gone through the roof, and I think a lot of retailers are battling to really understand: where is there shrink happening? So of course, the natural progression in that argument would be to say, well, self-service is a natural place for shrink because it’s unmanned in a lot of environments.
But what we’re actually finding is: There’s two different types of people that steal. There are people that maliciously try and steal and those that have just made a mistake and it’s genuinely unmalicious. And how you treat those two individuals has to be dealt with very, very differently, because you don’t want to alienate or embarrass the consumer that’s genuinely made a mistake.
For those that are maliciously trying to steal: unfortunately, if we close all of the loopholes and make it impossible to steal at self-service, they will find somewhere else in the store to go and steal. So, we’re in this kind of Whack-A-Mole-type environment, where we’re trying to close all the loopholes as quickly as possible. We’ve really focused our efforts on AI with behavioral tracking. And the reason why we use behavioral tracking is once you can start to identify behavior, it doesn’t matter where you deploy the technology within the store, you can identify malicious behavior and that shrink environment.
We very much focus on the frontend first: we’re deploying shrink onto self-service checkouts and onto POS lanes. But the idea is that the next natural evolution is that then run that same solution onto the CCTV network, and then we can identify shrink anywhere in the store. The human element of this is really important because it’s relatively easy to identify if someone has stolen. What you then do in that scenario is a difficult situation.
What you don’t want to do is alienate a consumer that might have non-maliciously stolen. If they’re maliciously stealing you also need to deal with that in a particular type of way, but you also don’t want to put your staff, your cashiers in your store: A, in danger; or B, in an environment that they don’t feel comfortable with. So, we are very much putting the human element back into this, that, depending on the use case of the theft, we will then deal with that situation differently.
But what we will always do is put the information in the members of staff’s hands so that they can deal with that situation in the way that they see as appropriate. So with all of our shrink solutions—whether it’s on self-service checkout or POS—once the shrink instance has been identified, an alert is sent to a member of staff’s wearable technology—whether it’s smartwatch or tablet or phone or even their POS lane—they’re notified that there’s a shrink instance that’s happened, they know where it’s happened, and they can even review the video clip.
So now they’re empowered that they know what’s happened in that situation, they know what to look for. And then staff training really comes into play here, and we have a number of great partners that we work with on staff training who actually work through these scenarios to give the staff members the toolkit so that when they approach that member of the public and they’re approaching them knowing exactly what’s happened, they’re trained to be able to deal with that situation in the most agreeable way possible—to disperse any aggression or any risk that might be associated, but also to make sure it’s a good experience for that end consumer. So, the technology is only one-third of the actual solution; the human element is a massive part of it that shouldn’t be overlooked.
Christina Cardoza: And I think the change in roles and responsibilities for cashiers to being able to have more meaningful interactions with customers—that’s not only benefiting the customer experience, but that’s also benefiting the employee experience as well, maybe keeping employee retention. I was a cashier in college, and I can tell you that is a tedious and redundant process. I would have dreams of just scanning food and shouting out numbers. And it’s not only retail shrink and loss—I think it’s not only with malicious actors or by accident—but sometimes as a cashier I would hit the wrong number just because I was on autopilot going repeat. It was an error-prone process. So, I can see that helping it as well.
You mentioned that to really be able to be successful you need an AI solution that connects all of these together so that this is not happening in silos and the data is actually actionable. Obviously, we’re talking to Diebold because you guys are a leader in this space. So, I’m curious to hear how you are helping customers—if you have any real-world examples or case studies that you can share with us.
Matt Redwood: Yeah. And I’ll be completely honest: we fell into the most obvious trap, looking back at our journey on AI. We’ve been working on this now for two and a half years, nearly three years. And we originally, we went out to market to try and find the best solutions to solve these three use cases, but what we quickly found were there was lots of different competing technologies. There were a lot of potential third parties that we could have worked with, but the underlying technology was the same.
And we quickly realized that actually as a solution provider who retailers work with to actually build out their technology—not just across their checkout but all the way across the store—it was unrealistic to think that we could have 20 or 30 different solutions, all in the AI space, all providing different use cases, but none of them talking together.
So, we actually kind of paused our program and redesigned our go-to market strategy, which was very much focused on providing an AI platform, and we work with a third party in this space who have a very, very mature AI platform. We’re entering the retail market and didn’t necessarily have the applications to run on top of it. So, we’ve worked with them to actually develop out these three applications as a starting point in the AI space.
But the nice thing about the AI platform is it effectively becomes the AI backbone for anything the retailer wants to do within their store from an AI perspective. This means that we can really satisfy our openness. It’s an ethos that we drive in our product strategy, which is openness of software. And what we mean by that is we provide the building blocks for retailers. We are the trusted partner, we’re the integration partner, but if there’s a particular third party out there who has got the market-leading solution in a particular area, it doesn’t make sense for us to go and reinvent the wheel.
So, when we talk about openness, the ethos that we take to our retail customers—but also that permeates through our R&D and product-management ethos—is to very much work with the best of breed within the market. And our strategy is to basically provide this AI platform for retailers. We will provide applications that can sit on top of it—like age verification, shrink reduction, item recognition, process or people tracking—but if there is a particular partner out there that is market leading in health and safety, we can plug them on top of the platform.
And what that means is the retailer can build this kind of ecosystem of AI partners, all providing best-of-breed solutions, but, critically, they’re all plugged into a single platform. So, they utilize the same business logic; they utilize common databases, like item database or loyalty schemes and things like that. That makes the solutions very, very scalable. It makes them much easier to manage, but it also means that they’re all talking to each other.
And the beauty of AI is it is self-learning to a certain extent. So, the more applications that we plug into this, the more physical touch points that we have in the store, the more information is flowing through the platform and then the quicker it can develop and the quicker it can learn. So, it’s very much a self-perpetuating solution that we’re very much at the beginning of this journey.
As I say, we’ve got about 54 different customers using AI in one form or another. But we very much see this as a much, much longer journey, where we’re starting to build an ecosystem of solutions that will ultimately move us towards what we call “intelligence store.” And intelligence store for us isn’t necessarily removing the physical touchpoint or removing the technology; it’s about providing intelligence to retailers.
And what I mean by that is every device that sits in the store is effectively a data-capture device. And that’s a two-way street: you can push data down to them, you can pull data back. So, whether it’s a shelf edge camera, or whether it’s a staff device or a self-service checkout or a scanner or a screen—these are all data inputs. There might be AI point solutions associated with them, but the AI platform allows you to connect all of these together and create an intelligent store, where intelligence really permeates every single area of the store.
It does mean there’s a huge amount of data available, but I think the retailers that are really going to advance quickly are the ones that work out what to do with this data. Because it can and it should inform every single decision or direction that you take as a retailer—whether it’s how I price my products, where my products are positioned within the stores, how I afford loyalty systems to the consumers, how I staff my stores, how I operationalize them—but also what technology exists within the stores.
So, data—it’s a cliché—but data will form the basis of every single decision that we make—from either a technology perspective, solution-provider perspective, but also from a retail-operations and a store-design perspective as well. So, it’s a really, really exciting journey that we’re on.
Christina Cardoza: Absolutely. And I think it’s really important to find a solution provider that is willing to work with others in the industry and leverage their expertise. I think that helps prevent vendor lock-in; it allows you to take advantage of the latest technologies and enables you to innovate faster, working with some of the best partners in the market. So, speaking of best in the breed, insight.tech and the “insight.tech Talk,” we’re obviously sponsored by Intel, so I’m curious if there’s anything you can tell us about that partnership and the technology that you use to make some of your AI retail solutions happen.
Matt Redwood: Absolutely. So, we work very, very closely with Intel—not just on the AI topic but from our core platform itself. Intel very much underpins a large part of our portfolio, so we have a very, very close working relationship with them—not just on the solutions that we deploy into stores today but also our roadmap on our development. We work very closely with Intel on their developments: where they’re going with their solutions and how we can better integrate them into our solutions to give retailers better solutions but also much, much greater flexibility for the future.
And I think a good example of that is probably the speed of development of technology. If you think about traditional point of sale or self-service checkout, if you go back five or 10 years, a retailer will make a choice for that particular type of technology and that would sit in that store for five, seven, 10 years sometimes, as long as the technology is running. The speed of development of technology is in increased immeasurably. The expectations of consumers have also increased immeasurably. And so balancing those two is really, really key.
Where we work very, very closely with Intel is on some of their scalable platforms. So, knowing that retailers have a requirement today—but particularly with these AI topics—the amount of computing power that will be required in three or five or seven years will be very, very different to the requirements today. So, providing retailers the ability to scale this technology so that whatever they deploy today is not throw-away in two years’ time. That they can evolve it and scale that technology to meet their technology requirements at that particular time is an absolute game changer. And that’s something we’re working with Intel very, very closely on.
Christina Cardoza: Yeah, absolutely agree. Things are changing every day: not even five, six, seven years from now, but five weeks from now things can be completely different. So being able to scale and to adapt is especially important in today’s landscape.
Well, it’s been great hearing about all of these solutions, especially how Diebold is helping retailers from end to end with the item recognition, facial recognition, and retail shrink. We are running out of time, but before we go, Matt, I’m curious if there are any final thoughts or final takeaways that you want to leave our listeners with today.
Matt Redwood: I think there’s a lot of misconception, particularly around AI. What I would say is: start with the data. Identify the business requirements or the problem that you are looking to solve, and then find the right provider that’s going to enable you to deliver against those requirements today but also gives you that longevity of scalability. Because AI is a journey; it’s very much a solution that learns over a period of time. So, choosing your solution provider is extremely important, because it is a marriage, and it is a long marriage, and you have to make sure that you’ve made the right choice. So, use the data to help inform those decisions, and, yeah, it’ll be very, very exciting to see where AI and retail technology goes, over the next two, three, five years.
Christina Cardoza: Yeah, absolutely. And I would say also: choose a partner that you can trust and is transparent about how they are using the data. Like with the age verification for instance, you want to make sure that that data isn’t being saved or that anything going into that—that system is going to protect your privacy and your information.
Matt Redwood: Absolutely. Data privacy is absolutely key and is a very, very careful consideration when you are designing or choosing the solution that you want to deploy to stores.
Christina Cardoza: Excellent. Well, thank you again for joining us. I invite all our listeners to visit the Diebold Nixdorf website: see how else they can help you in the retail space. As well as insight.tech, where we’ll continue to cover partners like Diebold and the latest trends in this space. Until next time, this has been “insight.tech Talk.”
The preceding transcript is provided to ensure accessibility and is intended to accurately capture an informal conversation. The transcript may contain improper uses of trademarked terms and as such should not be used for any other purposes. For more information, please see the Intel® trademark information.
This transcript was edited by Erin Noble, copy editor.
