DEVICE:
WEB
CLIENT:
AMAZON
YEAR:
2024
EXPERIENCE:
UX STRATEGY
about.
Designed a behavioral-nudge system that gave a two-sided logistics marketplace a lever to shift demand away from capacity bottlenecks — without any pricing or policy change — by testing what information, and what tone, actually moves professional decision-makers.
my role.
I led design for the 0-to-1 concept phase and the experimentation phase — from systems mapping through shipping a messaging system that fed data for a fully automated 2025 experience.
challenge.
Inbound shipment volume into the fulfillment network had grown 242% during peak periods versus a few years earlier, with the majority of that volume now coming from decentralized decisions made independently by thousands of individual sellers. Operations teams had no way to influence when or where those decisions landed, and existing warning language risked damaging trust with the very people the system depended on.
process.
Ran two workshops with 5+ cross-functional stakeholders (logistics ops, engineering, transportation, and adjacent product teams) to map operator personas and surface use cases, producing 66 insights and 16 divergent concept directions.
Designed and iterated 3 controller concepts against real technical sequencing constraints — the backend logic couldn't be reordered, so every design had to work with an existing decision sequence rather than around it — producing 100+ mocks across 12 review cycles in 4 weeks.
Identified that a planned messaging approach risked reading as punitive and eroding trust; escalated this to the responsible product team, who changed direction to test three separate message framings instead of shipping the original trust-eroding alert.
Designed the winning approach: transparent, factually-framed signals (e.g., "this region is near capacity this week") rather than directive or penalizing language, then partnered on A/B testing the behavioral effect.
result.
Framing one region as near-capacity moved 5% of shipment volume away from it versus a control group — validating that transparency alone, with no financial incentive, changes seller behavior.
Flagging specific weeks as congested pulled 7% of volume out of those weeks in advance versus control.
The approach scaled to 90% of base users within two quarters, then to 95% for a targeted peak-season rollout.
Once automated, the system hit record-low delay rates. On-time-shipment delays were cut from ~100% during congestion (pre-automation) to 2%, when 20%+ of operators shifted their region away from congestion after given visibility upfront. Furthermore, proactive delay-related contacts volume was reduced by 5%.