Walmart – Generative AI Comparison

Enabling customers to find the right item that meets their needs faster

We know that comparing items on Walmart is difficult, especially products with complex specifications. By curating summarized item information side-by-side, we boosted buyer confidence, saved customer time, and increased conversions.

+0.00%

Add to carts per visitor, indicating reduced friction in shopping decision

+0.00%

First-time buyer conversion, indicating easy learning curve

-0.00%

Session cart removal, indicating higher quality of add to carts

+0.00%

of customers re-engaging the tool, indicating high value prop

Role

Designer

Domain

Search, GenAI

Date

Q4 2023

Responsibilities

Discovery, Journey Mapping, Strategy, Research, Testing, UX, Execution

Publications on this work

Transforming Holiday Shopping With AI at Walmart

Oct. 30th, 2024 – 4 min read · Walmart Corporate News

Read article →

WalmartCorporate News

Item comparison on Search, powered by GenAI

Mar. 30th, 2024 – 2 min read · Jon Alferness, VP and CPO, Walmart US

Read post →

inLinkedIn Post

Discovery

Customer problem – comparing is high effort

Walmart customers find it difficult to compare products and make a decision online, especially for high-consideration items (priced at $100 or above).

Additionally, high consideration items make up 45% of shopping sessions across platforms

Along with benchmarking studies and voice of the customer feedback, we've seen comparing details between products being a huge pain point for customers.

Three TVs side by side representing the comparison challenge

Other competitors have a more dedicated way to store items on Search and compare

We are lagging behind many competitors, who offer an easier way for their customers to compare items across many different categories.

REI

Best Buy

Staples

Target

lululemon

Amazon

flipkart

Strategy

Guiding principles for comparison

Make it discoverable

Make the tool easily discoverable so customers can start to engage and develop a habit

Keep it simple

Reduce cognitive overload by keeping the comparison interface clean and uncluttered

Demystify terminology

Use natural, easy to understand language when explaining technical and complex specifications to customers

Be objective and unbiased

The tool should not favor any product/brand/supplier and be transparent about its methodology and how it generates results

Why now

Leadership wanted to further invest in large language models to improve and simplify the end-to-end digital shopping experience for customers.

Design

I explored several high level initial concepts

Standard chart

Knowing that a chart is industry standard, we wanted to innovate beyond this concept

Blog post

Although comprehensive, this design seemed too dense and resembled our item pages too closely

Input driven criteria

Data quality issues prevent us from leaning into asking for customer preferences

Summarized cards

Leadership loved the idea of summarizing crucial details for customers in one single viewport

Hypothesis

By providing recommendations with curated information, we will increase buyer confidence and reduce back and forth between search and item pages.

Research

Does this reflect how customers mentally evaluate choices?

My contributions: While consulting with my Design Researcher, Jen Luong, I conducted a DIY study which included drafting a study plan, conduct unmoderated study of the prototype, synthesis of results, and creating the insights/recommendation deck.

Takeaways from research

Entrypoint to the comparison tool was moderately discoverable

Many participants did not mention the compare nudge when evaluating the TVs, though many did quickly see it when asked. Those who didn't see it, said it's because they were focused on evaluating products.

Icons associated with the negative review mentions were confusing

Many participants looked to the tags to help them make a decision. The tags helped participants think about their needs/lifestyle. Several participants initially interpreted the "X" as a "No..."

Natural language titles reduced the comparison tool's effectiveness

Different titles added more effort: Comparison was quicker when the key differences titles and descriptors were similar amongst items

The AI-generated use cases were very helpful in narrowing options

Many participants knew what they'd typical use the TV for (movies, gaming, etc) and where to place it (living room, patio, bedroom, etc) so the highlights sped up the process. However, some participants didn't bother read the supporting sentence underneath.

Design

Buttoning up the designs

Responsibilities: Character limits, spacing, interaction, flow, error states, variants, AI-disclaimer with legal

Results

Design impact

+0.00%

Add to carts per visitor, indicating reduced friction in shopping decision

+0.00%

First-time buyer conversion, indicating easy learning curve

-0.00%

Session cart removal, indicating higher quality of add to carts

+0.00%

of customers re-engaging the tool, indicating high value prop

Next steps

After iOS launch, we're scaling towards more platforms and touch-points: desktop and cart

The team behind the magic

👤 Franklin Huynh (UX Designer)👤 Anisha Arora (Product Manager)👤 Yashwant Modi (Program Manager)👤 Keshav Agrawal (Product Director)👤 Charlotte Passot (UX Manager)👤 Shannon Lamb (UX Director)👤 Colin Mahan (UX Content Design)👤 Jen Luong (Design Researcher)👤 Shadab Ahmad (Discovery Engineer)👤 Vinoth Anandan (FE Engineer)👤 Vinay Hosamane (FE Engineer)👤 Shailesh Jain (BE Engineer)👤 Jason Cho (P13N)👤 Paul Yan (IDML)👤 Ronen Aharony (Aspectiva)👤 Sambhav Gupta (DCA Analytics)👤 Stephanie Miller (Business)👤 Himanshu Tanwar (UGC)👤 Ashish Parikh (Ads)👤 Priya Ramadoss (DCA Data Eng)👤 Anoop Saini (Catalog)...and many many more!

Up next

Nordstrom • Stores & Services • 2023

↑ 1.6% conversion↑ 24k new alteration users (1 mo after launch)↑ $1.4M EBIT profitability over the year