Case 04 · Amazon Robotics · RXD

Mapping exceptions and problem-solve handling across FC workstations.

A cross-station research framework for preventing, resolving, and routing problems at the source — built across two FC sites and five station types.

Wide hero shot of an Amazon fulfillment center floor — ToteASRS storage and active workstations

Role

Senior UX Lead — RXD (Robotics eXperience Design)

Team

Cross-program research, product, ops

Timeline

2025 — multi-quarter

Problem solve has been deprioritized — and it's compounding.

Across robotic workstations, exception handling is consistently moved downstream — never resolved at the source.

Every station creates problem solve work. None of them are designed to resolve it.

Why this matters now

Two upcoming workstation programs cannot inherit the current "figure it out" approach to exceptions.

New Pick-Pack station

Limited footprint, no sidelining room, blocked sightlines into the work cell.

Grocery workstations

Refrigerated and chilled items cannot sit idle on the floor waiting for a waterspider — this is a food-safety issue, not just an ops inefficiency.

Systemic, not isolated

Problem solve is a cross-station issue; solving it station-by-station won't scale.

Active safety hazard

Blocking station entrances with sidelined items isn't an ops inefficiency — it's a safety event waiting for a regulator.

A connected research arc

Rather than evaluate a single workstation, I built a five-stage research framework to trace problems from origin to resolution.

01 Identify What problems are being created and where?
02 Trace Where do items go after a problem is reported — or not?
03 Shadow How do PS associates resolve issues today?
04 Map Document the current exception process end-to-end.
05 Propose Develop workflows that reduce PS burden at the station.

Scope: 5 station types · multi-site · multi-method (shadowing, real-time observation, item-journey tracing, UX evaluations).

What it looks like when PS compounds

The cost of deprioritizing problem solve is visible at the station: totes accumulate faster than waterspiders can clear them, and a single PEW associate can surround their station in under thirty minutes.

PS totes accumulating around an inventory-removal station

Observed

25 PS totes from a single associate in 30 minutes — and 17 totes surrounding a single station at peak. The signal to clear them never reached the right person fast enough.

Operator at a station with sidelined items piling up

The current exception process — end to end

Four actors, multiple failure points, no one in control. I mapped every break in the handoff between operator, system, waterspider, and PS associate.

Associate Reports + waits
Missed scan signalsilent PVM (physical-virtual mismatch) created
System / UI Provides feedback
No reporting mechanismitem sidelined unrecorded
Waterspider Collects + transports
Waterspider dependencycollected only when noticed
PS Associate Resolves manually
One tote at a timeno bulk resolution

The swim lane became the team's shared artifact for understanding why station-level fixes never closed the loop — the breaks in the handoff between operator, system, waterspider, and PS associate were structural, not cosmetic.

How complex it is to resolve one problem

The PS associate's reality: no guided workflow, no prioritization — just tribal knowledge and six disconnected tools. Triage means stepping through up to six checks before landing on one of seven branching resolution paths.

Scan tote Check overheight Check weight Check disposition Check LPN Follow branch
Overheight PVM (physical-virtual mismatch) Missing expiration Expired item Unscannable Routing issue Mixed disposition

A senior PS associate built her own Google Doc to navigate this — six years of tribal knowledge. A new hire has none of it on day one.

PS associate's workstation — Google Doc reference, handheld scanner, and multiple tool windows open side-by-side

What can be solved at the station?

Before any recommendation, I filtered every problem type by whether it could be prevented, resolved at the station, routed, or only physically sidelined.

Problem type
At station?
Recommendation
Damaged / leaking item
Must sideline
Physical handling required — no digital fix.
Unscannable barcode
Partial
Print barcode at station OR re-induct to PS.
Unknown / unrecognized item
Yes
New guided UI workflow.
PVM (physical-virtual mismatch) from missed scan
Prevent
Enhanced multi-modal feedback.
Catch-all sideline
Yes
Replace with guided exception resolution.

Not every problem needs a new feature. Some need prevention; some need routing; only a few need physical handling. This filter became the input to a four-tier resolution framework.

The PS Resolution Framework

I authored a four-tier hierarchy that now guides every new workstation design — from new Pick-Pack to grocery.

01

Prevent

Stop the problem at the source. Enhanced multi-modal feedback, clearer scan confirmation, forced acknowledgment of failure states.

02

Resolve

Fix it at the station. Guided workflows for unknown items; barcode printers for unscannable items; replace catch-all sideline with structured resolution.

03

Route

Move it efficiently. Automated routing, consolidated PS totes (cap 1 → 11 items), batch-resolve workflows for PS associates.

04

Sideline

Physical handling only. Damaged or leaking containers — last resort.

What "Prevent" and "Resolve" actually look like

The framework becomes design direction. Three of the five recommendations turned into in-flight feature work.

Enhanced scan feedback

Multi-modal: full-screen background colour change + LED/captron lighting + auditory alert, with forced acknowledgment before the workflow advances. Existing hardware, coordinated signal.

Unknown item guided workflow

Exception button (footer, question-mark icon) launches a step-by-step UI that guides associates through identification — opened-item vs. unopened-case branches — using existing scan-error and confirmation patterns.

PS tote consolidation

Raise the per-tote cap from one item to eleven. Multi-modal PS alert (yellow background + red captron flash + auditory). Projected to eliminate 70%+ of station tote congestion.

Impact and forward adoption

The framework moved from a research output to an active gate on new workstation kickoffs and a roadmap for in-flight programs.

The largest outcome wasn't a UI; it was establishing that exception handling has to be designed before hardware ships — and giving teams a shared vocabulary for deciding how.

Next case studyMapping the Operations Tool Ecosystem →
Case 05

Get in touch

Currently exploring senior & staff-level UX roles in systems thinking and service design.