When people score an invitation to visit @WalmartLabs in San Bruno, Calif., Jeremy King says they often expect to find a small outpost of the $485 billion retailer, with 50 or so techies toiling away. Instead, they tour a vast, well-designed complex that houses about 2,500 people — part of a global team of 4,000 — who conceive, test, and deploy all of the technology behind Walmart’s websites and mobile apps.
King, the Senior Vice President and Chief Technology Officer of Global E-Commerce at Walmart, joined the company in mid-2011 to find an e-commerce division that was overly reliant on software vendors and on-site contractors for its technology development. He gave them six weeks notice, and began focusing instead on building his own team. “You can’t outsource innovation,” he says. “It’s something you really have to own.” (In 2013, he wrote a blog post that laid out his thinking: “Why Every Company is a Tech Company.”)
King was recruited to Walmart by then-CEO Mike Duke; he reports to Neil Ashe, the President and CEO of Walmart’s global e-commerce division. He spoke with InnoLead recently about some big ideas that have come out of the regular “hack days” that the lab runs; some projects that haven’t panned out; how the Agile development methodology has been bringing techies and business people closer together; creating a higher profile to help @WalmartLabs attract talent in Silicon Valley; and the vision that guides his group.
InnoLead: How do you see your role with Walmart, and how has it evolved over the years?
Jeremy King: The role we play here is, effectively, to push the boundaries of scale. We talk about all the things that are going on in the digital-physical world, with e-receipts and endless aisles and global payments and check-outs. How do I take that capability to a max scale for [more than 4,000] stores in the U.S., international, multi-language – all the types of interesting problems that the little retailers don’t have to worry about.
King: There’s about 2,500 at the lab. The entire division is about 4,000, so the bulk of the team is here in the Valley. And then I have a small team in Bangalore and another in Sao Paolo, and then a couple of other California divisions with about 50 people. Most people don’t know this, but we’ve been in the Valley for fourteen years. So we’ve just been expanding on our capabilities over the years here. Over the last four years, we’ve really changed the way Walmart thinks about technology.
At Walmart.com, we were basically an IT shop, where we bought all the software, used third parties to integrate them and shipped product. What I really did when I came in was transform the organization into an engineering team, where we largely built our own tech stack and effectively used open source technology — we not only use open source, but we also contribute back to open source. We run our organization like many startups … where I have many dedicated teams – over 100 of them that have their own KPIs (key performance indicators) and their own business partners – and their goal is to rapidly iterate just like a startup to develop new capabilities.
Keeping Teams Small
InnoLead: How many of those small, dedicated teams are there?
King: At last count, it was like 130. It varies up and down; they sort of spin up and spin down like startups do. There are teams for catalogs, personalization, the [product] recommendations engine … supply chain teams are broken up into replenishment teams and fulfillment teams and warehouse management teams. Each one of these has its own individual goals and objectives. We try to keep them under 10 people, so there are a lot of them.
King: One of the main things we had to do at the beginning was rebuild the tech stack. As I mentioned, we largely bought technology, so we had kicked off an effort two years ago to rebuild the platform from the ground up. That literally meant new data centers, new database architecture, new platform as a service, all the way through a new web framework. This effort, called Pangaea, was effectively building a global platform that all the different websites could use. As we developed those capabilities, we launched a new search engine, a new personalization engine, a new supply chain back-end, and a totally new interface for customers that is responsive and beautiful and has capabilities on the mobile phone as well as desktop – and lots of in-store features as well.
We’re doing in-store mapping and location. I know lots of retailers are doing that, but try doing it with 4,000 stores that are all set up differently. It’s quite different on a global scale. [Walmart has more than 11,000 stores in total.] This is one of the reasons we’re able to attract great talent here. Not only are we using the latest and greatest technology, but also our challenge is quite daunting. Engineers like to work on the hardest problems, right? Data scientists want to work at places where there’s lots of data, and believe me, we have lots of data.
Recruiting in Silicon Valley
InnoLead: Speaking of attracting talent, do you have to really go out and chase it, or are they lining up at your door?
King: I think it’s typical for Silicon Valley. There are a lot of options out there, and frankly, as busy as we have been talking about @WalmartLabs for the last four years, many people think of Walmart as the retail store. Even though we’ve been in the Valley for a while, many people don’t know we’re in the Valley. The good news is when they find out we’re here, we’ve been able to attract the talent – in fact, most [candidates] that we make offers to accept; it’s almost 85 percent, actually. (Here’s the @WalmartLabs recruiting site, which is separate from Walmart’s main careers page.)
We’ve hired about 2,000 people in the last three years, and we’ve also acquired 14 companies in the last three years. Mostly relatively small companies, but that helps as well.
Project Examples, Successful and Otherwise
InnoLead: Talk about some of the successful projects or strategies that have launched out of the lab.
King: The [Savings Catcher app] is a great one, where we took some data from our competitive pricing system and then we combined that with our customer e-receipt system and the mobile device. A customer scans their receipt after their transaction, and then loads it into their Walmart.com account. Then we’ll check prices for the next 72 hours, and if any of our competitors has a lower price, we just give them that money back in a gift card. It’s hugely popular for customers, as you can imagine. It’s directed to our hard-core customers, so they can save money and live better. There was very rapid iteration and delivery of that capability. (The Savings Catcher app, right, was released in August 2014.)
Another one we’ve been doing a lot more work on is wish lists and gift registries, in the online-offline channel. If I’m standing in the store and want to do a baby registry, could I also have online gift-givers? How do we integrate [things when the customer is] standing in the store and they want to scan, but they also may want to have options on the online site?
We just delivered a great new responsive [design] gift registry and wedding registry. Also, we’ve been testing grocery delivery in several markets. This is a very hot topic, and Walmart has had grocery delivery in San Jose and the Bay Area for about four years. We’ve delivered groceries in the U.K. for about 15 years [through Walmart’s Asda division], and we’ve expanded a similar program out to five other states around the U.S – and have been doing integration with the online site as well.
InnoLead: What about some projects that didn’t pan out, and what did you learn from those?
King: There’s one I’ll mention. When @WalmartLabs was first created, there was this big buzz about [social shopping.] The social commerce market – no one has really figured that out. Obviously, there’s Pinterest and Twitter doing their buy button and things like that, but in the end, social shopping has not exploded like people expected it would. So we played around with several different places where we were integrating with social media. There was a thing called ShopyCat that we did; we did another thing … which was essentially a social shopping thing where you could share your shopping ideas on Facebook and other social media.
We just haven’t found that killer app yet for social shopping. I think everyone thought social shopping was going to go huge.
Running Global Hack Days
InnoLead: What’s the process for generating new ideas and taking them through implementation?
King: Like most startups, we run internal hack days. So we just had our first one globally, where every single part of the site did a hack day at the same time – India, Brazil, and Silicon Valley. It was great. It almost turns out to be trade show kind of event where everyone comes in, and we do a whole day where people get to show off all the things they’ve built. Effectively, it’s 24 hours, but we cheat – we do the hacking on a Friday, and we do the demo on Monday, so that gives the teams a chance to do more coding. (Below, a team shows off their project to King.)
WMX, the Walmart Exchange, was a prototype from a hack day. This project allows people who buy digital media or digital advertising on the Web to show what the lift is inside a Walmart store, and this has been hugely popular with our suppliers. So if [someone is] selling a computer, and [puts] ads on Yahoo Sports and Maternity Day websites, I can tell them which ad had a bigger lift in a Walmart store, not even on Walmart.com. This is almost the Holy Grail for digital marketers, because they’ve been trying to justify their online spend for a really long time. But given Walmart’s share of market, I can accurately predict and tell our suppliers where their digital advertising is working and where it’s not. That was a great hack day project.
We also have a very traditional, Agile practice for doing project delivery – we use customer metrics and then we spec them out as business opportunities in a six-week cycle. We do that about three times a quarter – regular product reviews and all the things you would see in an Agile development [environment.] All the dedicated teams really help that. In fact, sometimes in [our] meetings, it’s hard to determine who’s the engineer, who’s the product person, and who’s the business person, because the systems person often knows as much about the business as the product and business persons do, and the business person knows so much about the systems now. This is why [structuring teams as] mini-startups is so effective.
Moving Away from Outsourcing Development
InnoLead: What are some of the challenges in connecting Walmart with the Valley?
King: Really changing the perception of what we we’re doing from a vendor software house to an e-commerce, engineering shop. That’s really the challenge, and frankly that was part of the emphasis for the @WalmartLabs name change. Before, they just called it a Walmart hub out here, and frankly they weren’t really contributing to open source. Walmart has been very conservative about talking about its technology in the media for a long time, and when I came on board we changed all that.
When most people come walking through our offices, [they] wouldn’t be able to recognize which startup [they] were in. It’s very much an open environment. There was a lot to do to transform that, from HR policies to facilities to the technology stack that we use to the way we evaluate people in our development process. It was a fun first year to get that all straight. [King arrived at Walmart in July 2011.] Actually, we did most of that in the first six months. We had outsourced a ton of stuff, and I let go of all the vendors in one day – I gave them a six-week notice, and they were gone six weeks later. That was the bulk of our team – it was all contractors. You can’t outsource innovation, as I like to say. It’s something you really have to own.
Becoming the Best at Digital-Physical Integration
InnoLead: What is your vision for Walmart innovation, both in the near future and further down the road?
King: Our destiny is to build out the best omni-channel experience you can find – the digital-physical integration. When I first came to Walmart and went through the interview process, I was very cocky about saying I can build my own data centers and have this all nailed in the first year. When I got here, I found out that Walmart has an unbelievable amount of assets – 200 million people walk into a Walmart every week, right? – so we have a tremendous amount of data coming in. Not only that, but also the brand is super-strong and what the brand stands for is really strong. (At right, a @WalmartLabs team member tests a new photo printing service in a store.)
Walmart is advertised as the world’s most efficient distribution channel, and I can connect directly into that distribution channel to build out an excellent e-commerce fulfillment center. So after we got our e-commerce basics in line, we spent tons of time building a better, deeper integration with our stores’ systems … from store mapping or real-time store inventory to the ability to do pickups in the store in a seamless way — all the things you would expect the world’s largest retailer to do. And that’s really what we want to do — we want to be the best at digital-physical integration.