Workflow manager (labor allocation) | amazon
UX / Product Design, UX Research
THE PROJECT
Disclaimer: This is a project summary. full details are available during portfolio reviews/interviews. thank you for understanding.
Amazon Grocery is a $24 billion dollar business consisting of 3 major departments: Whole Foods Market, Amazon Fresh, and Amazon Go.
Every Amazon Fresh site has business metric trackers on shipment volumes, grocery storage, staff pick rate, and more. There’s daily revenue targets to meet, and many of those targets are automatically generated based on live data that refreshes every 4 hours, 24/7.
Managers used almost 30 tools and dashboards to monitor their operations and site performance. Data was often inaccurate. This led to managers having to re-calculate errors, costing the business time and money. The “Workload Manager” project was created to consolidate several tools into to one single UI that would become the source of truth for live operational data.
the problem
Situation = Amazon Fresh managers were using a combination of online and offline tools to calculate their site’s performance daily. Data was less than 50% accurate, costly ($140M) and time consuming to fix (1.5 hours/week).
Task = Consolidate operations, inventory, and sales data from 29 tools into 1 scaleable dashboard for efficient planning and managing site performance with real time insights.
Action = User research, wireframes, iterations, mocks, prototypes, feedback sessions
Results = $147.3M total in cost savings, and 523 hours total saved/week (across 55 sites)
TEAM & ROLE
Amazon Worldwide Grocery Stores & Tech (WWGST) departments involved:
Product
Program
Operations, International + Domestic
Engineering / Tech
Fresh Stores
Whole Foods Market Partners
Role: Lead designer
Responsibilities:
Refining product requirements, pain points
User research (monthly), feedback
Wireframes, mocks, prototypes, iterations
USER persona & STORY
I created a user persona named Josh to help understand this project better. This persona is based on a compilation of user research I did at multiple grocery warehouses, and feedback I received from various managers throughout the design process.
Josh is responsible for calculating his site’s performance and how much headcount he needs on the site every day. The headcount relies on factors such as:
shipment volume, type, temperature
inventory management, online orders,
How much ‘backlog’ of groceries his site has to process from the day before, as the team often does not get through all of the shipments that arrive from vendors.
user RESEARCH
I conducted research monthly with our core users: area managers, site leaders, and multi site leaders. Research included: in-person interviews, online presentations and feedback sessions, and in-person feedback.
Some questions we asked included:
Who organizes and plans shifts?
What time (or frequency) do you normally plan these?
What were the existing tools?
Which tools have been the most/least useful?
What metrics tell you you’re doing well?
How does performance or under performance affect planning?
SITEMAP
Refined requirements with my PM and made individual notes in Figma, according to product and business requirements. This is a standard part of my process for all projects.
INFORMATION ARCHITECTURE
This is how I translated it (while referencing the product requirements)
interface design > custom components
In addition to Amazon’s Meridian design library, a few assets had to be provided by external plug ins, such as the Victory plugin for data visualizations. I adhered to the Meridian design library and their guidelines where possible. But I also used some of the existing assets to make custom components, such as the Key Metrics summary. This was to address 2 user pain points:
Old Excel templates offered a ‘big number overview’ that managers found useful. But this visual style was missing in the “new software” created for them.
Many managers expressed “feeling lost in the sea of data.” I learned from the research that every site had different needs, and what was considered a ‘success business metric’ for one, might not be relevant to another site. Being able to customize what they needed to see, and quickly, was essential to daily operations and monitoring site performance.
FINAL DESIGN
Interactive prototype link (Figma presentation)
RESULTS
Financial impact
$37 million dollar savings and $110 million dollars in ops = from improved labor scheduling and tracking
$174 thousand dollars in logistics leadership cost savings = from managers submitting tickets
$117 thousand dollars in cost savings = from staff intervention due to manual process
Time impact
91 mins per week
8 manager hours saved/site (assumed per week)
User impact
Manager experience (simplification of multiple dashboards, more control, greater satisfaction)
Reduce inefficiencies along the supply chain with manual process
Labor optimization (better SLAs, performance, quality targets and shrink)
Deprecated tools and tech migration that helped the business get closer to automated processes