Streamlining workflows to address $30M+ in inefficiencies and cut unresolved defects by 50%

Sideline App is where item defects get resolved after receive. When it breaks, every team downstream feels it, and the cost compounds fast.

V2 forced associates to bounce between Sideline, FCResearch (manual researching tool), and other internal tools to resolve a single item defect. V3 consolidates the workflow and cuts the unresolved rate in half across 10,000+ associates worldwide.

V3 was a full redesign across multiple features. This case study focuses on the two that had the most direct impact: the Scan Item page and the Verify page.

I owned this from research through worldwide launch. Facilitated the design sprint, flew to fulfillment centers, built the prototype, and walked engineering through every branch. I also pushed for two scope additions that shaped the final outcome more than the original brief did.

Every process path in a fulfillment center can surface defects. When they do, they get routed to Problem Solve. Sideline App is the tool PS associates use to work through those defects: wrong FNSKU, wrong count, damaged units, mismatched ASIN. While Sideline App touches many parts of the network, V3 was focused on the post-receive paths: decant, stow, and pick.

The V2 flow didn't match how defects actually surface. Associates jumped between Sideline and three other tools, lost context every time, and hit dead ends on edge cases.

Two moments did most of the damage: Scan Item surfaced no container contents, and Verify was missing essential item attributes.

In 2025, V2 and V3 ran simultaneously across pilot sites. V3 was opt-in for Problem Solve associates, a deliberate choice to collect early feedback and keep them in the loop from the start. We didn't push it. Associates who tried it told others, and adoption spread through word of mouth on the floor. By September 2025, the worldwide launch was met with positive feedback, and associates who never saw it coming adopted it naturally.

I led a one-day design sprint to align on V3 strategy, then went looking for what the sprint couldn't tell me. I shadowed associates across 13 fulfillment centers. Three patterns kept surfacing:

The foundation came from field research conducted in 2023. I picked it up, mapped the pain points, and used it to help the PM shape the Business Requirements Document (BRD), contributing research findings, design directions, and the core goals that became V3's brief.

The central design challenge was cognitive load. The Problem Solve process path already carries more context-switching than any other workflow in the building. Adding fields from FCResearch into Sideline App wasn't automatically the right move. We had to understand what information associates were actually looking at, why they needed it at that moment, and how it helped them make the right call to resolve a defect. The goal wasn't more data. It was the right data, at the right step.

V3 redesigned multiple parts of Sideline App. The two features below addressed the core breakdowns: Scan Item gave associates no context about what they were looking at, and Verify was missing the attributes they needed to make a confident resolution call. Both gaps pushed associates out of Sideline and into other tools, compounding the time and error cost of every defect.

If V2's biggest tell was that every associate switched to FCResearch immediately, the fix was to bring FCResearch's most-needed data into Sideline itself: catalog image, product title, ASIN, item quantities, and item status, all surfaced at the moment of scan.

FCResearch pulls much more, container history, inventory details, and broader research data, because it's built as a research tool. Sideline App isn't. The goal was to surface enough context to make the right call and reduce the need to go to FCResearch altogether.

My first iteration carried over V2's structure with more attributes layered on. Testing killed it fast. PSs were scanning the page looking for information in an order the layout didn't support.

During prototype testing and observations, we also noticed associates switching to FCResearch mid-flow, scanning the ASIN to pull the Amazon.com product listing. They were doing a virtual-physical mismatch check, confirming the item in their hand matched what the system was showing. It happened most on higher-value items where the stakes of a wrong call were higher.

Both signals pointed to the same problem: the page wasn't organized around how PSs actually work. I restructured the IA to match their mental model, and pushed for the Amazon.com link to be built into the product name so associates never had to leave the app to get that confirmation.

Problem Solve, like much of warehouse work, is a mix of software, hardware, workstations, and people in the loop. To get meaningful signal, we needed to test as close to the real experience as possible.

I pushed for and built a conditional logic prototype in Figma that mimicked the warehouse flow. Problem Solve associates could pick up any item, scan it, and move through the workflow the same way they would on the floor. Static screens couldn't validate something this branched. The prototype could.

V3 launched worldwide in September 2025. Across 10,000+ associates, the consolidated workflow delivered measurable gains in efficiency, accuracy, and trust.

Most projects at Amazon Fulfillment Technology don't follow a traditional agile process. For V3, we tried it, and it proved to be effective. As a UX team, we worked ahead of engineering, building and testing prototypes without pulling in engineering resources for front-end work. Engineering could focus on preparing the backend systems in parallel, using our mockups as the shared source of truth between UX, engineering, operations, and PM. Once designs were finalized, engineering was able to build and get everything ready before the pilot, which gave us the opportunity to catch and address bugs before the worldwide rollout.

Having a direct feedback channel and going onsite made a real difference. Being on the floor, watching associates use the product in their actual environment, gave us signal that no research session could fully replicate. It kept our iterations grounded and made sure what we were building was hitting expectations, not just meeting requirements.

The original goal was to remove FCResearch from the workflow entirely. We didn't get all the way there. Mental models and tribal knowledge built up over years don't disappear because the screen got better. What we did do was significantly reduce how often associates needed to leave Sideline App, keeping them in the workflow for the majority of resolutions.

On a personal level, this project pushed me in ways I hadn't expected. Leading the design end to end, bringing work directly to associates on the floor, and building the kind of feedback loops that actually change a product. It's a different kind of UX work, and it shaped how I think about design in complex operational environments.