Uber Eats is an online food ordering and delivery platform and a subsidiary of Uber Technologies, Inc. I was a member of a four-person team exploring the causes of delivery errors. Our redesign centered on three processes in the Uber Eats courier app with the goal of providing up-to-date and relevant information and reducing delivery errors. The result of our project was, according to Uber Project Manager Ryan Waliany, “the foundational piece of research that will inspire our humans-in-the-loop strategy at Uber Eats going forward.”

I worked with researchers in planning, conducting and analyzing user research. I coordinated with a second designer to build wireframes and prototypes in Sketch and inVision, conducted user testing, created a product demo video, graphic assets for the final deliverables, and presented the solutions to Uber.

As per an NDA with Uber, some project details have been intentionally omitted or obfuscated.

Uber’s internal Sensing, Inference and Research team had identified a pattern of missed deliveries or “errors” between couriers and eaters during the end stage of delivery. According to Uber’s sensor and machine learning (ML) data, the courier and eater were within 50 meters of one another, and yet the delivery was not completed.

What is causing deliveries to not be completed? How might we use existing sensor data or machine learning to alleviate these errors?

UX Designer & Researcher

• Tools: Sketch, inVision, Illustrator, InDesign, Photoshop
• Timeline: December 2017 - March 2018 (3 months)
• Deliverables: Design specification document, product wireframes, interactive inVision prototype, research findings, product demo video, presentation poster

Through our research, we discovered that couriers face many difficult situations throughout the delivery process. One area that was consistently identified by couriers was collecting orders from restaurants. Upon presenting our initial research to Uber, we requested that we redirect our attention away from the “last 50 meters” and focus on creating a better order pick-up experience. We designed new user flows and interactions within the Uber courier app which addressed top concerns surfaced by delivery drivers, including: parking selection, the in-app order pickup process, and a lack of support and transparency from Uber. Furthermore, our solutions streamlined existing processes within the Uber courier app, making the delivery process more efficient.

“The capstone team created the foundational piece of research that will inspire our humans-in-the-loop strategy at Uber Eats going forward.”

— Ryan Waliany, Project Manager, Uber Sensing, Inference & Research Team

“We [already] collect information on sensor and GPS data, but utilizing couriers helps us understand what’s actually happening on the road. Where they've parked, where they’re waiting, and using that information helps us make more accurate predictions on ETA and dispatch decisions. The idea of crowdsourcing from couriers can help provide guidance to other couriers making deliveries, [thereby] making this whole process more efficient. Even saving just one minute can lead to $50-100 million in savings at Uber's scale.”

— Ryan Waliany, Project Manager, Uber Sensing, Inference & Research Team

At the beginning of the project, my team and I conducted a competitive analysis of other food delivery apps, looking for common processes and patterns, and growing our knowledge of the industry. We reviewed materials from Uber depicting the “life of an order”, ML generated error reports, and previous research. Uber was unable to provide us with a complete user flow of the courier app, so we all signed up to be Uber Eats couriers. After completing several deliveries ourselves, we were able to stitch together screenshots of every step of the process into a complete master user flow, and gain a better understand our users. We interviewed research scientists at Uber to better understand the role machine learning and artificial intelligence will play in future product iterations, and how we can make affordances for it in our work.

In total, we conducted more than a dozen research interviews with food service workers, eaters, and couriers (including two courier ride-alongs). I personally facilitated three courier interviews and led one observational ride-along. We learned a great deal from speaking with couriers, from a lack of trust regarding Uber's business practices, to finding parking, and incorrect information in the app. Couriers are fiercely independent and incredibly resilient. Most often, when encountering an error, they figure out a workaround. Rarely do they report these bugs, due in part to a lack of reporting tools, but also a lack of confidence in support staff, and the cumbersome and time-consuming nature of the current support system. Couriers care about efficiency; making as much money as possible in the shortest amount of time. Time spent on the phone with support affects their bottom line. With that in mind, we also had to be conscious of the fact that any changes we suggested to the courier app could affect not only their experience, but also their livelihoods.

Perhaps the biggest takeaway from our research was that the majority of problems couriers face aren't in the "last 50 meters". Occasionally they do miss a delivery, but their frustrations are centered around the process of picking up orders and the experience they have with restaurants. When we presented our findings to UBER and they agreed to let us redirect our focus on the food pick-up experience.

My co-designer and I began the design process by developing a shared visual style that we would use in Sketch, based on the current Uber app. We both sketched numerous wireframes outlining possible solutions for each of the three focus areas and presented to them to the team and later to Uber employees for feedback. After discussing the various design options, we came to a consensus and began crafting an interactive prototype for user testing.

I led the updating of the order pickup and questionnaire wireframes, updating placeholder images and using as much existing written copy from the Uber app as possible. It was important to transition from wireframes to interactive prototype as quickly as possible, so I decided to utilize the inVision app. While inVision lacks complex interactions and animations, it provided us with a clickable prototype for usability testing in the shortest time possible.

Through social media, and our previous user interviews, we recruited six Uber Eats couriers to participate in a series of 45-minute usability testing sessions to evaluate our prototype. I worked with the researchers on the team to develop the testing script and conducted two testing sessions myself. The user feedback was invaluable and allowed us to address several problems with our prototype, such as users not understanding the heat-map style overlay, and unclear interactions on the order screen.

Following iterative changes resulting from user testing, we presented our project to the team at Uber. We not only provided them with our design solutions, but also demonstrated where ML can be used to customize the information presented to couriers, and how couriers can validate ML models through a humans-in-the-loop process. Our proctor at Uber presented our research and designs to several project groups within Uber Eats. The redesigned layout of the order pickup screen was well received, and elements of the design will feature in future product updates. The wait-time questionnaire, as a tool for gathering information and validating ML models, aligned closely with the work of another internal project team and will be folded into their development cycle. Unfortunately, providing parking suggestions to couriers could pose a liability problem for Uber and will not be pursued further at this time.

• Alexandra Olarnyk
• Janet Ng
• Audrey Wu