UX Design · Service Design · YOGO Robotics · 2021
Redesigning the in-building robot delivery experience for YOGO Robotics — reducing anxiety, simplifying interactions, and creating a unified mental model across robots and smart cabinets for delivery workers.
01
In-building Delivery was one of YOGO's main businesses. YOGO provided hardware and software to building owners, so delivery men (DMs) could call robots to deliver food to upper floors — saving time for both delivery men and end users. The system contained robots, intelligent storage cabinets, and a WeChat applet for both delivery men and users.
In this project, I worked as product manager and UX designer, aiming to improve efficiency and accessibility of the put-in experience and reduce anxiety for delivery men.
02
The system connects delivery men outside the building with end users inside. Each building has 2–15 robots and several smart boxes, depending on the building's height. When a delivery man enters a building, he scans a QR code and opens the WeChat applet to call a robot or cabinet. The robot then takes the lift and delivers packages to users.
This project mainly improved the experience on the delivery men's side.
03
"I'M GETTING LATE, SHIT!"
Being anxious of losing control was the main emotion we observed on site. Imagine a middle-aged man with little tech experience who needs to use a robot that "speaks nothing" to deliver successfully within 5 minutes during busy lunch hours — if he fails, he might lose an entire day's earnings.
In this case, any inconsiderate design would cause great trouble. We mapped out the user journey and pain points on using the robot delivery service.
04
We analyzed the previous design and identified the main issues creating obstacles for end users.
The system auto-assigned robots or cabinets without asking DMs. There was no visual feedback during loading, leaving DMs with no sense of what was happening.
When the system got stuck, it kept loading with no alternatives. DMs couldn't recall a mistaken package. Failure cases (robot breakdown, elevator unavailable) had no feedback.
Opening a robot required typing a code number, while a cabinet opened via a button tap on phone — two completely different mental models for the same action.
05
Over 70% of delivery men were new to a building — the service faced lots of new users every day. Two main personas emerged: one open to new technology (quick learners), another more passive. As the passive group was larger and faced more problems, we focused on them.
"I hope I could earn as much as I could for my child. Quick and right are the only things I care."
Wang Qiang came from Anhui Province. He moved to Shanghai to earn more money for his son studying at university. The busiest time was 11:00–13:00, when he had to rush to be on time. If he failed, he risked user complaints and lost money. He preferred simple shelves where he could just put packages and leave, and held a passive attitude towards technology.
06
Quick & Confident was the goal for the new design. We approached it through three main solutions.
The system proactively detects and fixes failures beforehand. For unavoidable issues, DMs receive clear reasons and next actions. In-process status is visible at all times, and DMs make the first step to choose which service to use.
We aligned the mental models of using robot and cabinet into one unified 5-step flow: select robot/cabinet → open door → put in → close door → leave. A Ring button helps DMs locate robots with sound feedback.
DMs can call back a robot to correct mistaken packages, cancel or change service type, and close the door via phone, robot screen, or by hand.
07
QR codes are placed high above cabinets or on robots to create a smoother moving flow.
Orders are auto-downloaded from the delivery platform. DMs choose between robot or cabinet.
Tap Ring button on phone to locate the robot with sound feedback.
Close to the robot, DM taps button on phone to open the door — same interaction for both robot and cabinet.
The robot detects movement and closes the door automatically with little effort required.
DM can call back the robot and reopen the box via phone to recover mistaken packages.
08
We prototyped and tested among colleagues, then ran an A/B test on some sites. The results confirmed that letting DMs take the first step to call a robot significantly reduced cancellation rates. We also ran on-site observations for quick iterations.
Many failure cases arose from building operations, technical limitations, and order overload. We clarified each case with clearer communication and pre-emptive detection.
09
According to observation and on-site staff feedback, the new design was easier for delivery men to learn and covered more corner cases. However, some challenges remained unsolved:
If a user didn't close the door and a delivery man put packages in without a system record, the only solution was to call staff to enter management mode manually.
Robot arrangement decisions were made on cloud, considering the whole building's elevator and people flow. This caused technical delays and occasional mistaken package assignments — beyond UX control.
DMs creating orders manually by typing the last 4 digits of phone numbers were very likely to make mistakes during busy periods.
This project taught me that a system with multiple terminals is complex and requires an even bigger supporting system. Some solutions are purely technical. The only direction is cooperation rather than thinking alone.