Multinational retail corporation that operates a chain of hypermarkets, discount department stores, and grocery stores.
How the world’s largest retailer reduced their cost of manually merchandising eCommerce products by over 52%
Walmart is the world’s largest retailer and #1 on the Fortune 500 list, generating over $485 billion in revenue annually. Walmart’s 2.3 million employees work at 11,000+ locations in 28 countries to sell consumer products and groceries at discount prices.
Walmart dominates in retail sales but struggles to compete with eCommerce companies like Amazon and Alibaba. Online retailers have less overhead and can carry 10's of millions of more products than Walmart’s brick and mortar stores. To better compete with retailers born online, Walmart eCommerce’s goal is to sell every product made by every brand and manufacturer in the world on Walmart.com.
Manufacturers and brands around the world have created 100’s of millions of products, but each brand stores product data in a different format. The process of standardizing product data and preparing a product for sale online is called product merchandising, and Walmart automates much of the merchandising process using a rules engine, machine learning, and artificial intelligence.
Walmart’s automated merchandising process enabled them to list millions of new products for sale online each month, but some high grossing products still required manual edits by the 300+ data entry workers on Walmart’s merchandising team. The old tools for making manual edits to products forced data entry workers to have 3 to 4 browser tabs open, and the tools couldn't edit every product attribute. Because of these limitations, the cost to manually merchandise a product was high, taking 36 minutes on average. Walmart needed a better tool to increase productivity.
Walmart’s internal team lacked the design capabilities and development bandwidth to create a single page application capable of significantly decreasing the time required to merchandise a product. After evaluating a handful of vendors, Walmart selected CognitiveClouds because of our experience creating modern web applications with AngularJS 2.0.
Our first task was to design the end-to-end user experience while our app development team waited for VPN access to Walmart’s internal development environment. We worked with key stakeholders on Walmart’s merchandising team to design high-fidelity wireframes, enabling data entry workers to edit every product attribute and group product variants together.
Walmart has a mix of legacy and new API’s, which means there are multiple ways to implement a feature on the front-end. Many API’s shared with us had incorrect or non-existent documentation, so we worked closely with engineers across several departments to identify the correct API’s to implement. Because different teams at Walmart gave our engineers conflicting advice, it took time for our team to find the right people who could recommend the best API’s to integrate with the single page application we were building.
Walmart’s backend data validation process was asynchronous and took anywhere from a few minutes to a few hours to complete. Business users on the merchandising team wanted to see the edits they made to products appear instantly on Walmart.com, but this wasn’t possible because of how the backend process was architected. To get around this issue, we created a screen in the merchandising tool, allowing users to visually confirm that edits they made to products passed or failed the validation process.
Sometimes we couldn’t persuade Walmart to develop new API’s to enable us to build the product merchandising tool the right way. In one instance, we requested that Walmart develop new API’s, allowing us to make a single API call to get data about multiple SKUs. Since other teams at Walmart didn’t have use cases for such an API, we were forced to make anywhere from 10 to 200 API calls on the UI UX design, front-end as a workaround. To solve this problem, we submitted all changes to product groups, and product attributes to the asynchronous validation process at the same time and optimized the frontend application to make up to 200 API calls when a user clicked save changes.
A new feature request by Walmart required us to display the product details for the same product carried by Amazon within Walmart’s content editing tool. To build this feature, we mapped each product attribute on Amazon.com to Walmart’s schema of over 10,000 product attributes. We used a UPC to ASIN conversion API to build an Amazon product page URL. Then, we loaded the product page in a headless browser hidden within Walmart’s content editing tool and matched each Amazon attribute with the same product attribute on Walmart.com.
During load testing, we noticed some performance bottlenecks caused by the network and hardware used by the 300+ data entry workers in India. We made recommendations to move to a content delivery network (CDN) to eliminate the bottlenecks and speed up the application’s performance.
Over the course of the 6-month project, we worked with Walmart to design and build a tool that gave new and enhanced features to the product merchandising team. After going live, data entry workers on the merchandising team were able to manually edit a product in 17 minutes on average, which represents over a 52% reduction from the previous average of 36 minutes. With the same budget, Walmart’s team of 300+ data entry workers can now edit more than twice as many products in the same amount of time.