“Be merchants first, technologists second.” In Walmart Technology, we hear those words on a weekly basis, but we don’t always get the chance to put them into practice. At the Greenhouse event beginning on July 6, teams got to do just that.
Unlike a traditional hack-a-thon, Greenhouse presents one problem and teams design multiple solutions. Greenhouse utilizes a design-thinking process where innovation stems from the desirability, feasibility and viability of the solution. Although the event is ultimately a competition, it’s also a collaboration; the teams learn from one another and improve each other’s designs as they progress.
Five teams in Data and Analytics across Bentonville and Bangalore were tasked with creating an artificial intelligence system called “Sam” that can make actionable, real-time recommendations to store managers. “It’s a big store,” said Wayne Bolinger, director of Store Operations Strategy. “From a store manager’s perspective, you can’t rely on knowing everything going on at all times. We need to build a solution so you can know the priorities without digging through paperwork to get there.” Essentially, Sam needs to be like a CEO inside every store, helping store managers prioritize in the moment with the best solutions to drive sales.
Over the next three days, they used the expertise of subject matter experts and store managers both in the room and around the world to craft their solutions. They participated in “fast-feedback” sessions at the end of each day, where they presented their day’s work and got suggestions from the other teams.
“It’s not a competition; it’s an integration between the teams,” said Preethi Krishnan, one of the Bentonville team members. “How can we work together so the company can be better, and how can we make Walmart better by having all of these ideas integrated?” The collaboration between teams was important; although the teams had the same goal, they each chose different paths and faced various challenges.
For instance, Balaji Ramlingam’s team wasn’t sure which problem to tackle first, so they went to the experts: the store managers themselves. “We prioritized the problems from their expectations,” he said. “We came to know real-time what the store managers are facing… It helped us to understand the business aspect of everything. We’re used to developing web applications, and this—we really learned something.”
Other teams knew where they wanted to start, but weren’t sure how to get there. “The challenge was getting clean, valuable data—actionable data,” said Georgi Gospodinov. His team’s prototype relied on a graph database that combines machine learning with a human-based data source. “Once you get enough users [on the application], you have the pool, but initially, we have to populate the data ourselves,” he said.
After three days of hard work, the teams were ready to present. The solutions were varied. They helped store managers partner with the department managers and identified individual sales drivers. They offered recommendations based on everything from sales to the weather. The solutions compared data to sister stores, used graph databases for real-time suggestions and combined science and art to create intelligent checklists and creative merchandising suggestions.
The judges went to deliberate. The winner: the team who’d begun their presentation with one simple statement: “Listen to everyone in your company.” Their solution did just that. The team in Bangalore was led by Nitin Sareen and included Rahul Rout, Viraj Patel, Aditi Ramanathan, Nikesh Srivasatava, Omker Mahalanobish, Issac Mathew, Jeebankumar Swain and Pralabh Kumar. Their prototype consisted of time-based triggers and continuous learning, but also allowed users to learn from interactions with their peers and subject matter experts.
“We were inspired by the self-driving car and how we could put our store managers’ collective wisdom into an auto-pilot learning mode,” said Issac Mathew, a data scientist at GTS. “Instead of going after point solutions using machine learning, we built a learning framework that would enable anyone who built such solutions to plug them into our framework. It was like building a university instead of designing courses. We blended a family of machine learning algorithms along with real-life experience from store managers’ feedback.”
Because of their remote location, the Bangalore team faced timing and technical challenges, but those only seemed to fuel their drive and commitment. They even stayed in the office until after 2:30 in the morning to present their prototype. “As the Data and Analytics team, we work on many initiatives across the retail lifecycle, and this is where it all comes together. It is the true value of analytics solutions—helping the store managers make the right decisions that help the business and customer,” said team leader Nitin Sareen.
“If a team of eight working for just three days can conceptualize and create a ‘wow’ solution, imagine what we could do if we take this forward with a larger group of passionate associates,” said Nikesh Srivastava. And that’s exactly what we plan to do. Leadership has committed to put together a team with 90 days to create the application, integrating ideas from all five participating teams to build the best possible solution for our store managers.
“The first Walmart Greenhouse was a brilliant example of how quickly you can drive game-changing innovation around real business problems by bringing technology, the business and users into a structured and fast-paced environment. We’re delighted with the results,” said Philip Black from Emergn, who hosted the event.
Now, we need to challenge each other to keep up the momentum; the talent and innovation within Walmart Technology will change the future. “We’re on the leading edge,” said CIO Karenann Terrell when she addressed the teams. “Take a minute to recognize that the ledge you’re standing on is history for Walmart… this will change how operations are done. When it’s done, it will change Walmart, and it will change Walmart Tech.”