UX research · Jan–Mar 2026
Mission Bay Shuttle
StatusResearch study · proposed directions, not yet tested
A free ride. An unclear first trip.

- Problem
- The free shuttle works, but first-time riders can’t find the stop, pick a route, or trust the arrival time.
- What I did
- As one of three researchers, I recruited riders, ran 6 interviews and a heuristic review, and synthesized 9 themes.
- Result
- Riders liked the ride but stumbled before it: reaching information online, finding the stop, trusting the time. Three proposed directions, not yet tested.
- 6
- rider interviews
- 30 min
- per interview
- 9
- recurring themes
- 3
- proposed directions
Problem
The ride is fine. Getting to it isn’t.
A free weekday shuttle connects San Francisco’s Mission Bay with transit hubs. We focused on students, employees, and residents making an unfamiliar trip.
- 248,507
- people rode the free shuttle in 2024
- 20
- stops across 4 routes to choose between
Research approach
We recruited riders in Mission Bay.
6 interviews, 30 minutes each, in person and on Zoom, plus a heuristic review of the website. Six interviews show direction; they can’t say how common each issue is.
Recruitment
Meet riders where they are.
A flyer in Mission Bay and a short participant screener brought in riders who had actually tried the shuttle.

01 · Listen
Six interviews.
In person and on Zoom, 30 minutes each.
02 · Connect
Nine themes from the notes.
Affinity mapping grouped the observations into nine recurring themes. Three shaped the work.
03 · Review
A heuristic pass on the site.
Nielsen’s heuristics, applied to the shuttle website alongside the interviews.
Findings
Three findings, one thing to keep.
Digital
5 of 6 hit a digital barrier.
The information was hard to reach.
Three different barriers, counted together: a hard-to-navigate site, relying on third-party apps, or not knowing the site existed.
Led toTask-based information architectureSee the decision ↓First trip
2 of 6 struggled on the first trip.
The stop was hard to find.
Physical signage and finding the stop.
Led toIdentifiable stopsSee the decision ↓Keep
3 of 6 liked the ride itself.
The ride works.
Safe, friendly drivers, and uncrowded buses. This is what any change has to keep.
Led toA design constraint
What riders told us
Interviewees described using alternatives to the official map.
Riders described confusion with the map and turning to other apps or asking someone for help.
What riders told us
Some didn’t know the shuttle existed.
A free service, invisible even to people who walk past its stops.
Paraphrased from rider interviews.
Keep what riders like about the ride. Remove the guesswork around it.
Recommendations
Three questions before boarding, one answer each.
The heuristic review found reachable schedules and a clean interface, so these proposals target specific gaps rather than a rebuild.
Recommendation → next observable task
- Find the stopShow stop name, routes, hours, and an arrival-information link togetherCan a first-time rider reach the correct stop without asking for help?
- Pick a routeGive routes, stops, and schedules one entry point with consistent labelsCan riders find their route and operating time? Record errors and hesitation
- Trust the timeDistinguish current, delayed, and unavailable arrival informationCan riders distinguish the states? Confirm data feasibility with the operator
Why it matters: in a Seattle study, riders with real-time arrival information said they waited 7.5 minutes on average, against 9.9 minutes for riders using schedules. Source: Watkins et al., Transportation Research Part A, 2011.
What I learned
I trusted the riders over the audit.
A research study, so no launch data. What stuck with me:
A combined theme can inflate a finding.
“5 of 6 hit a digital barrier” merged three different problems: navigating the site, relying on third-party apps, and not knowing the site existed. Now I report themes separately, even when the bigger number is more persuasive.
The ride wasn’t the problem.
Half the riders described safe, friendly, uncrowded trips. What people like is a design constraint too, so recommendations don’t break what already works.
The audit and the interviews disagreed, and both were right.
The heuristic review found reachable schedules and a clean interface, yet riders still couldn’t find or trust them. A usability review needs people doing the real task beside it.
Unreliable data needs an honest state.
Riders distrusted the tracking, so I proposed showing current, delayed, and unavailable states instead of a live prediction. When data can be wrong, how its status is shown comes before a better guess.
What’s next
Test next
- Real trip-planning tasks with first-time riders
Check with the operator
- Whether the arrival data can support current, delayed and unavailable states
