Problem definition

Switching back and forth between two devices was time-consuming

• Friction in user workflows: As we move functionality over to Nexus, Inventory Control associates are required to carry both the legacy and Nexus devices and switch back and forth between the two. They cannot be logged into both devices at once, so they have to take the time to log out of one device and into the other each time to switch between workflows.

• Inaccurate inventory data: 1k+ adjustment input errors per day were created, and suppliers could mistakenly be credited. 

• High engineering costs: The backend system was difficult to work with from an engineering standpoint and deploys are cumbersome.

Measuring success

Building frictionless experience to facilitate accurate inventory data creation

Our team relied on the success of these metrics:

1. Percentage of accurate inventory data

2. Number of warehouses using this tool (Adoption)

3. Engineering developer velocity

Analyzing the existing workflow

Three flows were complicated and unclear to associates

Three flows on the existing old tooling were confusing to users on which workflow to pick. Three flows were:

• Scan Location

• Scan License Plate (LP)

• Scan Item

New workflow

Consolidating three flows into one simple flow

I thought it was the opportunity here for us to consolidate three flows into one single flow that could optimize the adjusting process. I explored different options based on the usage data we found:

• ~68% of total adjustments were done by scanning LP

• ~ 55% of total adjustments didn’t require to scan item

We landed on a flow that allowed users to scan Location or License Plate (LP). Scanning item was needed only if users selected Multi Carton Kit (MCK) or Remap as adjustment reasons. The final flow looked something like this:

Design iteration

Adjusting interactions

I explored different interactions on how to adjustment in/out inventory QTY. Our usage data suggested that ~90% of inventory adjustments were done in bulk instead of single carton movements. Although -/+ were the typical buttons on a calculator, such interaction would be slower for adjusting a large amount of items. So we decided on the option where users could type the number directly on the keyboard.

User acceptance testing, surprising data

Missing an important feature that allowed users to conduct bulk adjustments 

During user acceptance testing (UAT), we uncovered that we had missed incorporating a feature that allows users to adjust multiple units without having to re-enter the reason and reference information each time.

I quickly scheduled and led a meeting with our teams to walk through 4 possible solutions. It would take our engineers more time to build two separate flows for single or bulk adjustments. And ~50% of our associates only used the single adjustment. We finally landed on this final solution where two options were displayed to support two uses cases: bulk adjustments or start a new adjustment. We quickly pivoted and completed this update which was released on July 1 (just one day after quarter end!).

Outcomes

“I pick up very quickly, and we have less chances to leave it wrong!”

• Successful launch: reached 97% adoption across 15 warehouses

• Better user experience: consolidated 3 inventory adjustment flows into 1 single flow; removed user reliance on memorizing 20+ adjustment codes

• Quicker development: 40% increase in developer velocity through Nexus framework

Reflections

Using data & qualitative insight to inform robust design

The question I asked most during this project was: hey, could we get the data on XXX? There are pros/cons from a UX perspective in both versions, so going with the data makes design decision easier and scientific. However, we always have to make tradeoffs when the data speaks something different from user needs.