Automating Fresh Distribution
Tesco, the UK’s largest grocery retailer, operates stores of all sizes, supported by a network of distribution centers. To reduce food waste and costs, Tesco launched a project to automate fresh product handling with autonomous robots and digital tools, improving accuracy and efficiency.
My Role
As a Senior UX Designer, I led Tesco’s “Project Emu,” aimed at improving efficiency in fresh grocery distribution centers through automation and digital products. This included integrating Autonomous Mobile Robots (AMRs) into workflows—the first initiative of its kind in the UK. My role focused on ensuring the technology addressed operational inefficiencies and fit seamlessly with existing processes.
Product Goals
Tesco faced challenges from labor shortages and wage inflation, driving up costs. To maintain affordability for millions of customers, they needed to cut food distribution expenses while increasing efficiency. Fresh grocery distribution suffered from poor inventory management and low pick accuracy, costing £100M annually to process 600M cases of fresh products. The project targeted a 35% efficiency improvement with KPIs such as:
Decant Rate: 518 cases/hour
Pick Rate: 325 cases/hour
Perfect Cage Accuracy: 99.5%
Cube Utilization: 26%
User Research and Analysis
Our team worked directly from the distribution centers, conducting daily site visits to observe, shadow, and interview colleagues in their workflow. Through these interactions, we captured key user roles, pain points, and motivations, focusing on how existing processes could be improved to meet performance incentives and work more efficiently.
Collaborating with stakeholders, we mapped out workflows to identify where digital solutions could enhance automation and support employee needs.
Testing included guerrilla, unmoderated, and live production environment trials, enabling quick iteration based on real-time feedback. This hands-on approach linked digital solutions to the physical workspace, giving us a comprehensive understanding of colleagues' challenges and allowing for designs that aligned closely with operational realities.
Conceptualization and Validation
I started by working with the default application provided by the automation supplier, which was originally designed for pharmaceutical clients. Using it during setup, I quickly realized it didn’t fit the physical workflows or needs of Tesco’s fresh goods warehouses. Testing the system with colleagues, I identified key areas for improvement and redesigned the interfaces for marshalling, decant, induct, pick, and loading. To tie everything together, I built a scalable design system to ensure consistency across the warehouse. I also created a manager’s app to help track the performance of both colleagues and robots, giving managers the insights they needed to make better decisions on the floor.
Pick Process Enhancement
Throughout implementation, I worked closely with our third-party automation supplier, Distribution Operations, and Tesco colleagues to test and refine early prototypes and interactions between the humans and robots. Spending 40% of my time on the warehouse floor was a valuable method of reducing the feedback loop and bringing users into the design process. This connection to users and the physical space they work within was crucial to building a complete understanding of the issues our users face, such as a lack of feedback from devices (e.g., reduced vibration and auditory feedback), which led to slower picking speeds.
To address these challenges, I developed two significant improvements:
The team measured and tracked site pick efficiency daily and early data analysis revealed that pick rates were far below the required KPI. Observational research showed that colleagues struggled to locate items on shelves brought by robots, as products often looked similar (e.g., Mature vs. Extra Mature Cheddar). Product information was presented in small, English-only text that did not meet accessibility guidelines. Collaborating with colleagues, we redesigned screens to show product locations as images, leading to a 16% increase in reaction speed and memory. Adding color coding between shelf and device further boosted identification speed by 40%. Following this, a quick implementation of color markers on shelves resulted in a 14% increase in pick rate within weeks.
Decant Process Optimization
Decanting pallets was efficient with one or two items, but most pallets had three or more, requiring colleagues to remove and re-scan items individually. To improve this, I collaborated with Operations and our automation supplier to enable a 3D view of pallets by scanning each side, allowing identification by both rows and columns. This new approach supported the induction of up to nine items per pallet, covering 85% of typical pallets and improving throughput by five times, which significantly cut down on pick time.
I frequently collaborated with colleagues between shifts to co-design and A/B test physical and digital solutions. For digital testing, we used laptops and arm-mounted terminals with scanners to simulate the production environment. Physical prototypes were tested in a mock station area outside the robot grid, allowing colleagues to trial concepts before full development.
Outcome and Impact
Through these targeted interventions, I helped Tesco achieve measurable improvements in efficiency and productivity:
Pick Process: 14% increase in pick rates through visual and color-coded designs.
Decant Process: Fivefold increase in pallet induction capacity, leading to faster processing and reduced pick times.
The project not only met its ambitious efficiency goals but also enhanced user satisfaction by making the technology more intuitive and aligned with colleagues’ real-world needs.
Lessons Learned
This project reinforced my belief that digital products are only half the solution—they must be designed in conjunction with the physical environment they serve. The collaboration between cross-functional teams, the empowerment to make decisions based on KPIs, and the importance of a shortening the feedback loop were key drivers of success.