No items found.
3
min read

Winning Customers with Algorithms: How Teams in Nairobi Help Shape Your Shopping Experience

Termed “planogramming,” visual merchandising is key in retail stores. The best stores find a balance between exciting customers without overwhelming them with deals shouting from every corner.

Full Name*

Email Address*

Phone Number*

Subject*

Topic

Thank you! Your submission has been received.
We'll get back to you as soon as possible.

In the meantime, we invite you to check out our free resources to help you grow your service business!

Free Resources
Oops! Something went wrong while submitting the form.
Winning Customers with Algorithms: How Teams in Nairobi Help Shape Your Shopping ExperienceAbstract background shapes
Table of Contents
Talk to an Expert

As a kid, I marveled at the strategy involved in retail layouts. From inviting department store entryways stocked with deals, to the tactical placement of milk & eggs in the backmost aisle of grocery stores, there’s always a reason why an item is placed where it is. It was a never-ending personal challenge to try and decode these invisible maps.Termed “planogramming,” visual merchandising is key in retail stores. The best stores find a balance between exciting customers without overwhelming them with deals shouting from every corner.Walmart is the reigning king of retail. With an average of 120,000 products sold in each store, visual merchandising is critical in maximizing both sales and profits. Yet, as with all traditionally brick-and-mortar markets, technology is re-shaping the consumer experience. Forget 120,000 products — eCommerce sites now place tens (and sometimes hundreds) of millions of products at consumer fingertips.A new visual experience is emerging, one that starts and ends at a computer screen. Think of Walmart.com as the equivalent 150 Walmart retail stores combined into one super-megastore. Nearly 1 square mile of product (think 2/3 the size of Golden Gate Park!). You land on the homepage and are surrounded by mile-long aisles, each with thousands of virtual shelves — where will you wander today?With these massive assortments, the key to success in emerging eCommerce marketplaces is ensuring that the consumer finds what they need. Whoever brings the most searchable shelf to the consumer’s eye wins the virtual planogramming competition.Keys to navigating this marketplace can be basic — when you search for a white t-shirt, you don’t want results to include a red dress. They can also be more complex and predictive — when you search for firewood, schematic search may also recommend you also buy lighter fluid. As a result, each product needs a significant chunk of metadata.Take a stroller as an example. Beyond knowing that it’s a baby stroller, what color is it? Fabric material? How much weight can it bear? Is is appropriate for a newborn? Does is require assembly? Is it a multifunction stroller?With a rapidly growing product assortment, data scientists at WalmartLabs partnered with Sama to create an easy, user-friendly way for consumers to navigate millions of unique items.These data scientists create predictive algorithms that automatically assign attributes to products, so that computers can predict with confidence that this is a pink and black stroller made with polyester fabric that can support up to 50 pounds.At its core, this data is inherently human. Behind each prediction, each “1” and “0”, is a judgment that was originally done by humans. Many of Walmart.com’s predictive models “learned” from training data supplied by our Sama team in Nairobi, Kenya.In one instance, Walmart created an algorithm to automatically predict the “gender” associated with products listed on their website. The goal was to sort products that had gender specific words in the title or in the description, such as “Nike Air Women’s tennis shoes” so that these items could be identified as products for women. Walmart partnered with Sama to create data sets for creating the ground truth, and training the algorithm. Sama manually QAed 115,522 products — this included checking to see if the algorithm had correctly predicted the gender, and if not, assigning the correct gender.As is typically the case prior to the creation of a training data set that provides sufficient ground truth, at the start, the algorithm’s accuracy was around 59%. After using Sama data sets to train the algorithm, WalmartLabs engineers were able to increase the algorithm accuracy to 93%.Walmart’s partnership with Sama has resulted in visible enhancements in product descriptions, therefore improving the shopping and purchasing experience of its extensive portfolio of products. Moreover, this project provides work for 12 Sama agents — talented women and youth in Kenya for whom access to dignified data work represents a path out of poverty.Next time you shop online and over-buy because you keep finding exactly what you want, blame our team in Nairobi! Or blame the internet for creating virtual shopping carts that never get full.

Author
Andrew Ho
Andrew Ho

RESOURCES

Related Blog Articles

No items found.