Walmart Reorder Recommendation

In 2023, the Ads business team approached Walmart My Items team to surface more sponsored recommendations to customers in the Reorder space. As the main designer on My Items, it was my responsibility to design an experience that would help surface the most relevant ads to the users without overwhelming them.

Ads Team Partnership

Walmart has a system of 4-in-the-box stakeholders between design, product, engineering and business. Our business partners are an important component to how we prioritize the projects on the timeline, so when the Ads business team approached the My Items to improve the ads presence in Reorder page, I got to work.

As profitable as ads are, it is a common annoyance to users, especially in a highly personal space like My Items Reorder, where customers come to easily reorder their previously bought groceries and consumables. Because of this, the ads that are shown in Reorder has to be as relevant as possible to the customers, and presented at the right times.

Initial Ideation

During ideation, I was inspired by the ‘clip strip’ strategy commonly seen in offline retailers. Stores often show items that pair well together in context of each other (i.e. putting Oreos next to the milk aisle or wine opener next to the wine). Naturally, a customer who came to buy one item would also consider buying the other, complimentary item (which also benefits the customers— what’s a bottle of wine if you can’t open it?).

There were 2 approached to showing these recommended items to the customers. One was the clip-strip strategy of showing complimentary items and the other was showing close-match items when an item from their Reorder is out of stock (Think customers going to get their usual olive oil from the store but it is out of stock— what do they do? Turn to the next row in the aisle to buy a similar product)

Explorations

With the two approaches, I iterated on a few different designs to present to my partners and leadership. One caveat was that, because our engineering resources are limited, we had to align on a relatively low-effort solution that could be implemented within the given timeline.

Complementary items

Similar items

AB Tests

Across leadership and engineering syncs, the team landed on the designs to be AB tested. The final design showed a similar item next to a Reorder item that the customer is used to buying, which was a mix of the proposed strategy from design.

The main metrics we were looking for in the test were increase in ads impression with no negativity to ATC (add to cards), which would indicate that the ads are not detracting the customers from shopping. In addition, we kept a lookout for increase in customers’ ATCs and basket size, which tells us the effectiveness of items being recommended.

Next steps

The test results proved to be successful with $7M annualized ads revenue with net positivity in ATCs. The design has since launched into production.

As next steps, the My Items team wants to test a modified version of the originally proposed designs where a mix of complimentary and similar items are shown at the end of each ‘aisle’.

However, the main user complaints from the original test was around the irrelevant recommendations. We wanted to make the users feel heard and also gather feedback around the accuracy of the algorithm. As a result, I enhanced the designs to help hide the module and remove recommendations as needed by the customers.

Reflections

This projects leaves a lot to be desired, still, due to the limitations we faced in terms of resourcing and business goals. There were many designs along the way that could have been a better solution to showing sponsored items that correspond to users’ needs but were turned down because it didn’t meet the right business requirement or were too high in engineering effort.

I hope that as My Items and Ads team continue their collaboration, we explore ways to monetization that put our customer needs first.