Walmart logo

Walmart Split Orders

Requesting an additional driver for large deliveries.

The stay-at-home order for COVID-19 across the nation has caused many to order in bulk for their grocery deliveries.

With a partnership with 3rd party delivery services, this has led to orders being too large to fit inside a driver’s vehicle.

I co-designed with my peer, Yangyang, on a new feature that enabled associates to split large orders by requesting a new driver.

Walmart's TC-70 device our associates use for fulfillment


Launching at the end of Q3.


What do I do if an entire order won't fit inside the vehicle?

Currently, for large orders that don’t fit inside the driver’s vehicle, some associates partially load them and manually call a new driver to come pick up the remaining items.

This wastes a lot of the associates' time and leads to customers wondering where the rest of their order is.

How might we provide a seamless experience for an associate to handle orders too large to fit inside a driver’s vehicle?

Images of Walmart's Dispense process
Walmart's dispense process (handoff to the customer or driver)


Uncovering the use cases

I first aimed to gather more insight into this particular problem and scenario. I conducted user interviews with our associates to learn more about what they usually do in these situations.

  1. Associates hack their own way of splitting up orders since there's no standard process.
  2. Most of them don't even let the customers know that their order has been split.

I worked with my PM, Santhosh, to summarize the different use cases we found and documented them on our Confluence page.

Table to capture different use cases
Different use cases

This page helped our team better understand how the issue is currently handled and enabled us to think of a more systematic approach to solving the problem.


Virtual whiteboarding to unlock our thinking

Yangyang and I did a whiteboarding session using Mural. We noted key associate and customer pain points as well as possible ways of addressing them.

The possible solutions we noted were intentionally forward-thinking and didn’t take into account development effort. We would later scope this out with engineering.

Our virtual whiteboard session via Mural
Our Mural doc


What if we could predict the likelihood of an order being too large in the beginning and prompt for the associate to take the action of splitting it.

Version 1 design flows

Having synced with our engineers early, we found that this approach wouldn't be feasible at this time because we would need a deeper integration with our third-party services (DoorDash, Postmates, etc).


Keeping the momentum going

The next 2 weeks were used for iterating on designs and conducting virtual usability tests. We held daily check-ins with our engineering and PM partners as well as weekly design crits.

Version 2 design flows

Summary of learnings:

  1. Associates couldn’t find the Split order action hidden within the context menu. They were confused on the primary instructions on the screen and weren’t sure what “Split” meant. They questioned the illustration because it's against their policy to leave totes on the ground.
  2. Associates didn't notice the instructions copy.
  3. The ability to check/uncheck totes and items was very well received. However, there would be too many systems to work with in order to get the item level details so our team didn’t move forward with this for the launch.
  4. Associates needed more reassurance that a driver has been called.

After 4 more design iterations with reviews and tests, we landed on the final flow for single orders.

Final version for the design flow


Single order

The associate splits one order into two trips at the time of Dispense.

Walmart's TC-70 device our associates use for fulfillment


Batched order

The associate splits a batched order into multiple trips at the time of Dispense.

Walmart's TC-70 device our associates use for fulfillment
Design spec documentation
Some design specs


Important data points to pay attention to

By the time we started crafting our final prototype, I worked with our PM and Data Scientist to come up with key metrics to track in order to determine the success of this new feature.

Success metrics table


The release of this feature got pushed to the end of Q3 because of a new strategy roll out plan from the company.

When this launches, our team will pay close attention to the metrics listed above.


If you try to uncover all the use cases, you'll never ship. If you never ship, you'll never uncover all the use cases.

I read this somewhere online and it really resonated with me. For this project, we were spending a lot of time trying to think of every possible scenario that might happen.

While these are very important, it's better to not dig too deep or you'll waste time and design for cases that never happen.

It's critical to launch and learn once it's live.