The Home Depot
Optimizing Human Effort
Internal Enterprise Mobile App
Enterprise UX Manager
I led the UX team responsible for Sidekick, partnering with Product and Engineering on near-term improvements and longer-term product direction.
01
The Problem
Sidekick is an enterprise mobile application used by thousands of Home Depot store associates to manage assigned work throughout the day. The app had many opportunities for UX improvement, but one of the most consequential went beyond the interface: the system had a limited understanding of the work it was creating and assigning.
Manager-created tasks were essentially text. People could understand what the words implied, but the system could not. System-generated tasks had more intelligence, including machine learning used to prioritize work based on selling patterns and profitability, but still had limited awareness of the realities of doing that work.
The result was inefficient use of one of the store’s most valuable resources: associate time.
02
A Task Is More Than a Sentence
My research exposed the gap between how Sidekick represented work and how associates actually experienced it.
A task might say: “Replenish HDX 27-gallon Tough Storage Totes in Aisle 10, Bay 12.”
To Sidekick, that was a task.
To an associate, it meant something closer to: that’s a large item, this could take 20 minutes, I’ll probably need to close the aisle and use a forklift, I need to locate the pallet first — and right now, I’m helping a customer.
The sentence was simple. The work wasn’t.
Sidekick had little awareness of the real-world implications of creating and assigning that task.
03
Modeling Work in the Real World
That led me to define a two-horizon strategy.
First, my team would continuously improve the Sidekick experience associates already depended on, measuring success through UMUX-LITE scoring.
Second, I partnered with Product and Engineering to explore a richer model of work that could eventually allow the system to reason about more than the task description.
A task could carry attributes such as urgency, duration, location, required skills, certifications, equipment and completion criteria. The system could also consider the associate: where are they, what are they qualified to do, what are they already working on, are they helping a customer.
This reframed the problem from displaying and prioritizing tasks to understanding work well enough to assign it intelligently.
04
Improving System Intelligence
With a richer model of work, manager-created tasks could eventually move from unstructured text toward structured data.
Instead of simply displaying whatever a manager typed, Sidekick could understand important characteristics of the work and use them alongside operational signals to make better decisions about priority and assignment.
That created a path toward a system that did more than tell associates what work existed. It could begin helping determine what should be done, when, and by whom.
Outcome
3.2 → 4.5
UMUX-LITE score
We executed against both horizons: improving the product associates depended on while defining what the next generation of Sidekick needed to become.
My team continued shipping improvements to workflows, information architecture and interactions, increasing Sidekick’s UMUX-LITE score from 3.2 to 4.5 out of 5.
At the same time, my research established a clearer model for more intelligent task creation and assignment. Implementing that direction required significant engineering investment, but the design problem was no longer ambiguous.
Shipped: UX improvements → 3.2 to 4.5 UMUX-LITE. Strategic work: research and a new work model → requirements and direction for smarter assignment.
Next case study
Taming Retail Chaos
Computer Vision Platform
Read the case study