Problem Space
Tourists are struggling
We reviewed 10 academic papers on tourist decision-making models, and conducted interviews with 10 travelers, uncovering 3 key problems they commonly face.
Information Overload
Trust and Credibility
Uncertain Conditions
Uber Drivers and Tourist Riders
When riders travel to a new city, they often crave authentic local experiences, but they lack personal connections. On the flip side, drivers know and love their city. They want to share it, but don’t have a clear or comfortable way to do so. Both sides want to connect, but right now, there’s no smooth or reliable way to make that happen.
Tourist Riders
Have no personal connections with locals of the city they are visiting
Know and love their city, but have no clear way to share it
Unsure whether locals are willing or available to offer recommendations
Worry that their recommendations may be unsolicited
Interacting with locals can be unpredictable, awkward, and rarely documented
Struggle to fully convey their recommendations while driving
Uber Drivers
Research Question
Aside from Airbnb, no existing platforms truly rely on locals for insights. Uber is uniquely positioned to do so through its network of local drivers.
Research Question
Aside from Airbnb, no existing platforms truly rely on locals for insights. Uber is uniquely positioned to do so through its network of local drivers.
Research Question
Aside from Airbnb, no existing platforms truly rely on locals for insights. Uber is uniquely positioned to do so through its network of local drivers.
Research Question
Aside from Airbnb, no existing platforms truly rely on locals for insights. Uber is uniquely positioned to do so through its network of local drivers.
Combined strategies are needed here
Three Themes
The three tourist challenges were translated into three themes for ideation. Each team member generated 10 design concepts inspired by research findings, and all ideas were categorized under the three core themes as shown in the Venn diagram. While many ideas successfully covered two themes, none fully encompassed all three. We realized that addressing them all would require a combination of strategies. Following the brainstorming session, each team member selected five design ideas with the most potential and developed sketches to better understand how these ideas could fit within the context of the Uber app.
Information Overload
Personalized
Trust and Credibility
Trust
Uncertain Conditions
Quick
Concept Testing - Round 1
We selected three concepts that were well-received by participants during concept testing round one. The first concept recommends points of interest (POIs) along the rider’s current route. The second reimagines POI selection through a dating app–style swiping interface. The third replaces traditional ratings and reviews with insights drawn from Uber’s unique trip data. While none of these concepts alone became the final product, each contributed critical elements to the overall solution.
Concept Testing - Round 2
Rider experience product flow: after gathering feedback from the first round of testing, we developed new sketches that began shaping end-to-end product flows. One flow focused specifically on the in-ride experience, mapping the full journey of a first-time user interacting with what would later become known as Uber Recs. This work marked the initial formation of the Uber Recs experience.
Driver experience product flow: we also developed a flow for the driver side, specifically within the driver app, exploring how local insights and recommendations could be prompted from drivers without adding significant burden. However, due to time constraints, the project scope was limited to focusing solely on the rider-side experience.
Final Concept
We translated the envisioned experience into concrete steps, procedures, and interface elements needed to support it. The flow diagram outlines the full user journey. This exercise helped identify the critical decision points, interface needs, and system triggers required to make the experience feel seamless and intuitive. With a clear understanding of both user flow and system behavior, we were ready to proceed to developing the high-fidelity prototype.
Current Ride-booking Page
Design Goal
Add Uber Recs into the flow.
Keep it simple and intuitive.
Iteration 1.0
Iteration 1.1
Solution 1
Each ride option (UberX, Black, etc.) includes a dropdown menu that lets users choose from nearby drivers.
Iteration 2.0
Solution 2
Display available drivers directly under each ride option, without adding a Recs mode toggle.
Iteration 3.0
Solution 3
All existing ride options remain unchanged, and users can simply choose drivers from a separate section called Uber Recs.
Open Uber at airport rideshare pickup
Book a ride to hotel with Uber Recs
Wait for pickup and view driver profile
Save driver recommendations
Approaching en-route recommendations
Book a ride with saved recommendations
Overall Process Summary
👥
Recruited 12 participants.
9 frequent Uber riders, 3 former Uber drivers.
📱
Prototype was presented on a smartphone at high fidelity; users were encouraged to think aloud.
🤔
Users were not introduced to Uber Recs up front which allowed for observation of organic discovery.
💬
Each session ended with a semi-structured interview covering usability and expectations.
Task 1: Discover Uber Recs
Scenario: Users arrive at Atlanta airport and are booking a ride to their hotel.
Objective: Observe if users organically notice and interact with Uber Recs without prior introduction.
Task 2: Converse with the Driver
Scenario: A team member role-plays as the Uber driver during ride.
Objective: Observe if and how Uber Recs enables natural rider-driver conversations.
Task 3: Act on Saved Recs
Scenario: Users decide where to visit the next day in their hotel rooms.
Objective: Observe if users can find saved content and book a ride to it.
Localness
Test Finding
Users were drawn to recommendations from locals.
10/12 users (83%) trusted recommendations more when accompanied by a personal quote or a photo of the driver. Many described this as “real” or “memorable.”
Experts warned that personal narratives might risk discrimination during driver selection.
Experts highlighted the need for ethical safeguards if rider preferences or driver attributes are surfaced too explicitly.
Curation
Test Finding
Fewer, clearer recommendations helped, but not all users agree on how few is few enough.
11/12 users (92%) described the 3-recommendation format as “digestible,” “low-pressure,” and “easy to remember.” One user called the list “weirdly limited.” “There’s no way those are the only cool places nearby.”
Some users found driver-centered recommendations too heavy.
They preferred general suggestions that didn’t require interpreting the driver’s background or taste.
Consensus
Test Finding
Users seek social consensus, but only when it aligns with their own gut feeling.
7/12 users (58%) said they trusted recommendations more when grouped under “# of Drivers Picked This.” However, 9/12 (75%) disregarded the most popular recommendation and questioned the consensus.
Users instinctively look for negative reviews to "balance the story."
4/12 users (33%) demanded negative reviews, even though they found the layered consensus helpful.
Spontaneity
Test Finding
Users welcomed recommendations that popped up mid-ride, until it felt like too much.
7/12 participants (58%) saved or tapped on a recommendation when it appeared just before passing it.
One user said the pop-ups felt “like a nudge I didn’t ask for.”
Another asked for more control: “I want to decide when I’m in ‘spontaneous mode’, not have it assume I always am.”
© 2014-2023 Yijiang Xu
SCROLL TO TOP