Uber Eats. Delivery robots, ten weeks of research.

What it takes to trust a delivery robot.

Research / Product · IDUS 215 · Contextual Research, SCAD · Spring 2026

Project details

Course: IDUS 215 · Contextual Research, SCAD
Year: Spring 2026
Duration: 10 weeks
Tools: FigJam, Miro, Next.js, Claude Code, Vercel

Brief.

What does it take for robotic last-mile delivery to be adopted with confidence by customers, couriers and restaurants? We studied Uber Eats and its Serve Robotics robots in the United States.

Context. Can people trust a robot at the door?

Ten weeks studying how customers, couriers and restaurants experience robotic last-mile delivery.

Question and setting

The question was what it takes for robotic delivery to be adopted with confidence. We looked at Uber Eats and its Serve Robotics robots in the United States.

The work was a ten-week contextual research course at SCAD, IDUS 215. It ran in five stages: planning, fieldwork and cultural probes, affinity diagram and synthesis, an ideal experience framework, and recommendations for Uber Eats.

Infographic. Four slides, from market to mechanism.

Conditions, system flow, user behavior and mechanism, one slide each.

What each slide says

Slide 1 sets the conditions: demand up 63%, robotic delivery growing about 150% a year, roughly 80% of delivery cost in driver labor.

Slide 2 maps six steps from order to pick-up and marks the friction in the last meter. Slide 3 covers trust and the emotional response. Slide 4 turns it into numbers, like one supervisor for every five robots and a +300% throughput gap.

Process. Six methods, checked against each other.

We combined desk research, probes, interviews and campus intercepts, and kept an insight only if two methods agreed.

How it was made and what was delivered

Two participants kept video diaries for five days. We ran 6 customer interviews and 6 restaurant interviews, plus sensory intercepts on the Georgia Tech campus. The diaries caught what interviews missed, like the half-second of hesitation before someone unlocks the robot.

Everything was coded into an affinity diagram and run through the Lextant framework to reach six emotional drivers.

We delivered a client book for Uber Eats: methodology, synthesized findings, the framework, personas, opportunities and next steps. It is built as a website with Claude Code and hosted on Vercel.

Live research site. The full client book, in the browser.

The research site is embedded here, loaded only when you ask for it.

Uber Eats delivery robots, research site

Open the site

What is inside

Navigate the introduction, methodology, analysis framework, ideal experience and strategic outputs, from research statement to next steps.

Learnings. Trust is decided at the hand-off.

Four findings carry the conclusions: the hand-off, the brand, the weather and the framework.

What we took from it

The two seconds after the lid opens decide trust. Food that arrives hot and intact buys forgiveness. When it doesn't, no notification fixes it.

Robots from different companies look alike, so brand identity is gone by arrival. A recognizable livery is a trust cue.

Weather breaks trust in interviews, and our campus intercepts happened in ordinary weather. That is a gap in our own fieldwork.

Rewriting observations as desired-experience statements made the findings usable. Each of the six drivers now has a feature and a sensory cue behind it.

The two seconds after the lid opens decide whether people order again.

Next project

Sillage.

Product · IDUS 100 · Project 6: Wearable Skin · Prof. Chunshik Kim · Spring 2026

A wearable fragrance diffuser built like a skeletonized watch.