The world of retail has changed dramatically over the past decade, and in ways far beyond a black-and-white shift from shopping in stores to shopping online. Today, e-commerce is table stakes, meaning companies distinguish themselves—among other avenues—via user experience, promotions, fast shipping, and omnichannel experiences that integrate digital and brick-and-mortar locations.
In this episode of the Cloud & Culture podcast, Shaun Anderson of VMware Tanzu Labs—this time joined by colleague Brandon Blincoe—concludes his introduction to the Swift methodology by focusing on retail application modernization. The two share experiences working with large retail clients and, in addition to the topics highlighted in the quotes below, touch upon retail-specific concerns such as pricing, inventory management, and aligning technology systems after mergers or acquisitions.
You can listen to the whole episode in the player. Keep reading for some highlights.
When customization eclipses off-the-shelf e-commerce platforms
Anderson: “The e-commerce tools, or anything that you buy off the shelf, by design it kind of has to be generic. You can't customize things so much because there's only so many things you can do with an off-the-shelf product. . . . Like, if cost of the product is $5.99, but companies that want to be competitive may say, ‘Well, I would like to apply a discount to people who are part of our loyalty program.’ So, the pricing of that product changes, and the off-the-shelf packages typically don't make it easy for you to do something like that.
“Or if you want to plug in some other kind of marketing, or you want to say, ‘Hey . . . a product may now become a bundle. So my product is the Little Gretzky package, which includes a hockey stick and a helmet and pads and skates.’ But those platforms often, just by design, can't support that because you can't do everything for everybody. So you just do the bare minimum.
“. . . There's usually a transition when retailers, or companies in general, start to get big enough where they see the competitive advantages to be able to be flexible. That breaks that model pretty quickly.”
Blincoe: “I think it's the ability for these companies to incorporate their brand into the fact that they're selling these products. They want to market it a certain way to their end customers. And I think, for instance, the hockey bundle is a perfect example for that. These platforms just don’t give the capabilities to do that easily and quickly.
“The other side of it is you want to run a weekend deal, sort of a last minute thing. The marketing team wants to do that and they see some huge upside in it, but the software package doesn't actually allow for it.”
It’s easier to hire for common skills than for product-specific experience
Blincoe: “The way that we approach looking at, for instance, an e-commerce process flow, there's many systems that interact with each other that make up that. . . . When we look at it from the lens of the business processes, we're not looking at it from, ‘This system is this technology, this system is this technology,’ and we can start to make decisions of how the system wants to behave, how the entire e-commerce process wants to behave.
“Some of the challenges that retailers are facing are around the fact that they are locked into these COTS products. They had invested in these products in the early 2000s, mid 2000s. It's really difficult to hire—just from a straight execution standpoint, it's a very specialized skillset. But if we now start to look at these blocks as business processes, we can start to think about how we might replace these with new technology and be able to hire for that new tech. It's a way to retain talent and bring in some new talent.”
Inventory management, technology choices, and what’s good enough
Anderson: “I'll tell you a little bit more about inventory, that's kind of interesting. There tends to be what I'll call a ‘technology overcorrection.’ . . . You'll see retailers, or anybody for that matter, who want to think in terms of, ‘I want to know real-time inventory. I want to know how many of these products I have at this particular given point in time.’ And a lot of times people will say, ‘OK, well from a technology-solution perspective, we could implement that with an event-sourcing tool like Kafka, and then they start going down this technical rabbit hole.’
“But with Swift, what we'll do is ask stupid questions: ‘So tell me, why do you need that?’ And often the answer is, ‘Well, I need to know when I need to reorder that particular product.’ And it's like, ‘Do you need to know exactly? Or is it fuzzy? Like when you get down to 10 percent, then you reorder. And if it's 10 percent, then you don't need to go through that extra dramatic expense just to have real-time.’
“And we even ask stupid questions when you're talking about the e-commerce versus the in-store [experience]. In e-commerce, if somebody puts something in their shopping cart, it's like, ‘OK, well, we mark that, but how long do we keep it? At some point they walk away, they don't buy it. Is it really available for sale again?’ Or ‘How do we tell if somebody's got something literally in their cart, in the store? It's not available for sale anymore, but our real-time inventory says it is.’ And then pretty soon people pull their hair out. When all you have to do is say, ‘Is it good enough just to know when you're close? '
“And that's similar to the fault tolerance versus high availability discussion. Do you want fault tolerance like it never breaks? Because that's an order of magnitude more expensive than just having resiliency and fault tolerance. Same story here. But you have to have a vehicle for asking those questions typically—where people are comfortable, too. Because it sounds like a dumb question, but it drives out a really good answer.”
The business case for Kubernetes in the store
Anderson: “[Kubernetes clusters in physical stores] is a reality and it can be a very good idea. As an example: Many large retailers essentially have data centers in each of their stores. The Home Depot, for example, Walmart—most of these companies have, in the back room, a rack of computers that are keeping track of their inventory locally, in large part because the infrastructure wasn't available to have something in the cloud. And it's really important that the cash registers are working, so you want to have that edge as close as you can.
“And in other cases, like a drug-store company, they have the need for edge where it's really just functionality in the store that's managing the security and the prescriptions, for example, that you don't want spread across multi-region.
“There are reasons I've seen it start out as edge and remain that way, and one is because there's so much traffic to keep track of—30,000 products on The Home Depot shelf—and it was much easier to implement something where we have what amounts to a Kubernetes cluster in-store that manages all of those different domains and capabilities, manages the cash registers and even the people and the scheduling and the inventory. And then maybe periodically phones home to the mothership to try to keep things in sync.
“In that example of ‘Do you really want real-time inventory or is close enough, close enough?’, this is a good example where you get that local resiliency or security and you can take advantage of the cloud for everything else. And even do things like, ‘We know that it seems like today we've got this many items, so we'll go ahead and sell them through e-commerce. But we may do the logistics through the stores and say, we don't have it at this store, but we can ship it there, and maybe not even let the customer know that there's some of that inter-store shipping going on.’
“And that wouldn't be as possible if you couldn't phone home to the mothership periodically, but more importantly have the computing power at the edge—whatever that means —to solve that problem for you. It's just a lot less risky.
Blincoe: “I think resilience is the key there—having the ability to have a store be self-sufficient and continue to run even if there's spotty internet, a power outage, that sort of thing.”
More on the Swift Method
About the AuthorMore Content by Derrick Harris