From paper to digital, a picking experience now used by thousands of grocery partners, from local corner shops to national chains.
Senior Product Designer, co-leading with a peer Sr. UX Designer
Challenge
A new digital experience to support pickers collecting Just Eat grocery orders that should support all grocery partners, from big national chains (e.g Waitrose, ASDA, Iceland, etc) to independent corner shops.
Constraint
Legacy restaurant-first product flows, limitations of Zebra devices, and tight delivery windows.
context
Grocery represented a significant growth bet for Just Eat, but attracting new partners or building partner adoption was being blocked by operational friction at the picking stage. Before the digital solution, Just Eat grocery orders were handled using printed receipts. Pickers manually cross-referenced items and tracked their progress, which made the process slow, error-prone, and disconnected from the order management system.
The brief wasn't to build for one type of store. It had to work for big chains (e.g., Waitrose with 40,000 SKUs) and a dedicated Zebra device fleet, as well as for an independent grocer with 2 staff members using personal phone. Research confirmed the complexity: chains needed speed and reliability, while independent partners needed to feel in control and connected to their customers. A single flow had to serve both realities.
In collaboration with UXR, we did In-loco research with store visits across London (chains and small/medium grocery stores)
DESIGN APPROACH
Each phase required a different approach: execution, orchestration, and iteration based on real user feedback, along with business and commercial needs.
Phase 1: De-risking the MVP in a highly constrained environment
In this phase, our design challenge was to create a new workflow that supports pickers in their core jobs while we navigate our legacy product, which is focused on restaurants and has outdated design components. We also need to consider the hardware limitations of Zebra's devices, such as low screen resolution, while adhering to a tight deadline and meeting the strict requirements of one our chain partners.
Phase 2: Defining the Roadmap & North Star
This phase was design-led, during which we synthesized all the insights gathered from our research and the previous phase to collaboratively define and shape the next improvements for the picker app with the product team. Our primary focus was on how to deliver impact by reducing the time spent on picking and enhancing order accuracy. In this phase, we also concentrated on refining the visual components and considering their scalability.
Item categories
To-Do / Done tabs
Camera scanning
Phase 3: Driving value and adoption throught substitution
Through data analysis and field research, we found that Out-of-Stock items were a significant issue, impacting around 12% of orders (up to 30% in chains) and causing considerable revenue loss. Our challenge was to create a solution that helped pickers in large chains find substitutes from 40,000 SKUs while keeping it simple for local corner stores using mobile devices.
We had two options: build an ML-powered recommendation engine, or ship something simpler faster. So, instead of developing a complex machine-learning engine from the start, we opted for a phased rollout. We focused on a 'Pre-picked list' for the most common scenarios and a 'Global search' fallback, allowing us to bypass legacy constraints and deliver value more quickly to partners, which also impact adoption of our solution.
REFLECTIONS
To positively influence the development of our product, we established a set of guiding mindset pillars. These pillars are thoughtfully crafted based on insights from research, user behaviors, and the specific jobs users need to accomplish. We also considered the context and scenarios in which they will engage with our system, ensuring our decision-making process is aligned and effective.
The UI must complement the physical environment, not compete with it. Glanceability and clear system status (audio, notification, visual cues) are more important than complex layouts.
Progressive disclosure (like splitting tasks into active/completed) is critical in high-pressure operational tools to reduce cognitive load and maintain speed.
In high-volume operations, mistakes cost money. We learned to remove friction from the "happy path" but intentionally keep micro-friction at high-risk moments to ensure accuracy.
Impact
Design decisions that moved the needle, measured in revenue saved, recovered, and protected.
Revenue saved through the Substitutions feature
Revenue recovered in a single month
Out-of-stock orders, now recoverable
