Amazon: Pickup Lockers



PROJECT OVERVIEW
Amazon Lockers are designed to provide customers convenience and security — so they never miss a package ever again. That’s only if the package makes it to the locker.

How might we ... ensure driver behaviors, software and hardware are working together to meet the needs of  customers?

THE IMPACT
  • My designs reduced defects per million by 88% (260 DPMO ︎︎︎ 31 DPMO)
  • I saved Amazon $387K in replacement orders and >$4.5M in operating costs to Amazon Lockers

*Based on 36M multi-family delivery volume in ‘22, 260 DPMO and ave cart size of $47. 


Why Amazon Lockers?

If you've ever lived in an apartment, you know it can be difficult to receive packages. They get lost, misplaced, or dropped in front of the wrong door.

In 2017, Amazon launched 3.5K lockers globally to meet this need. 


PRESS RELEASE
 

The Challenge

5 years later, Amazon Pickup was delivering 267M packages in 20 countries annually. With that scale comes certain operational and logistical challenges.

Two Challenges:
  • Near-term: 36M packages delivered to multi-family units, 2.5% were considered “Delivered Outside Locker”. Of those considered DOL, 260 Defects per Million occured due to “Package not being Detected”. 
  • Mid-term: as Amazon’s merchant network expanded, Amazon could send more types of package sizes (ex: super thin or large) and needed to reconfigure their locker slots to pair supply with demand.


We believe that in the near-term ... improving the UX inside the driver app would reduce 95% of package not detected defects.
We believe that in the mid-term ... reconfiguring the package slots would drive throughput and reduce Delivered Outside Locker by 0.75%


Why are these opportunities important:
  • Amazon saves $5M/yr in operational costs for every 1min of unnecessary action they eliminate in a drivers day (275K drivers). 
  • Locker deliveries are ~40 sec faster than home delivery (+/- 13 sec).
  • Lockers are a scale and cost savings mechanism (ex. 1 delivery = 50 packages [locker] vs. 1 delivery = 1 package [home])
  • Throughput is primary metric Amazon Hub is measured

Success Measures
Primary
  • x% reduction of DPMO due to ‘Package not Detected’
  • x% reduction of ‘Deliveries Outside Locker’ 

Secondary
  • x% increase in driver satisfaction at multi-family housing



What I Did

DISCOVERY ︎ DEFINITION ︎ DESIGN ︎ DEVELOP/RELEASE/EXTEND

DISCOVERY = 2 weeks
In practicing human-centered product development, I conducted just enough research to ensure the problem solved a real need. 

Includes: Service Safari, downtown & rural route with research report



My approach to research:
I test every week with customers. Internal stakeholders, a prototype, sales team or customers.
  • Usability Test detail designs right before release (A/B).
  • Evaluative Research lo-fi ideas, wireframe, or prototype.
  • Generative research (exploratory, foundational, discovery) the broad context of our customers.


DEFINITION = 1 week
Before hi-fidelity screens, I take existing data and visualize the problem, forcing myself to articulate the opportunity within our system constraints — in order to achieve faster outcomes.

Includes: information architecture, data, sketches, use cases 


Low-Fidelity Sketches:




Use Case for ‘Package Not Detected’:
  • 80% — envelope sized packages are not detected if they are not placed on top of the sensor.
  • 18% — package legitimately will not fit in the locker.
  • 2% — sensor is damaged and needs to be replaced by a technician.


DESIGN = 2 weeks
With an understanding of the need and the constraints, it was time to increase the level of visual polish — all within the Rabbit Design System and their components.

Design Components:



Hi-Fidelity exploration



Driver Testing
My team tested a number of prototypes with delivery personale during “Driver Tuesday” a standing session for interviews with drivers who use the app.
 
Topics included:
  • Group Stops
  • Package Not Detected




DEVELOP/RELEASE/EXTEND = 2 weeks
Because this was a B2B product the bar was lower than customer facing app, so we opted not to QA and just went right to product with my redlines. 




Mid-term exploration
In order to help the ‘Delivery Outside of Locker’, I mocked up several concepts to pass off to the fabricator who would build the new Lockers and expand our package capacity.




Challenges Encountered


You have to bring people along the journey otherwise its just waterfall strategy. I came into the project asked to build some screens without context. Work always happens faster when you bring people into your work. 

Getting access to drivers wasn’t hard, but finding routes that did Lockers was challenging. Goes to show there is always some element of hustle in finding your target customer.

CFx Partners
  • Tim White - UX Design
  • Joe Kraisorn - Product/Business
  • Arun Milton - Product/Tech
  • Garima Jhunjhunwala - Tech

©2025