This analysis examines 4.6 million BayWheels ride records to understand user behavior, operational constraints, and scalable growth opportunities in San Francisco’s bike-share network.
The core finding is a structural imbalance I call the Assymetric commute: bike-share is heavily relied upon for time-sensitive morning commutes, but far fewer riders use the system for their evening return trip. This creates station-level deficits that cannot self-correct and must be resolved through manual rebalancing.
For this project I analyzed BayWheels Data from 12.2024 to 12.2025.
Goal: Identify who and when use bikes to define core users archetype.
In the absence of user identifiers, I inferred human behavior through membership status and trip geometry.
I focused on:
-
Member vs. Casual split, using membership as a proxy for retention.
-
Station pairings, distinguishing point-to-point commute corridors from round-trip.
Key findings:
Casual riders are approximately 3x more likely to be riding for sighseeing or experience (round trips) than Members, who are likely using bikes for point-to-point utility or commuting.
Two dominant archetypes:
- Utility-driven commuters (members) ride for efficiency
- Exploratory or recreational riders (casuals) use cycling as a leisure
I used station-level return patterns as a proxy for churn.
Stations with:
- high casual volume
- low membership %
- high round-trip rates
were identified as experience-first location.
Examples:
-
West Crissy Field and Lincoln Blvd at Hoffman St show 11–13K rides, low membership (20–40%), and high round-trip behavior, consistent with sightseeing use.
-
23rd St at Santa Clara St and Saint James Park show high membership (~78%) alongside moderate round trips (16–22%), suggesting locals using bikes for recreational loops, not tourists.
Outliers with extremely high membership and round-trip rates (80–90%) were removed using the IQR method.
To measure station stability I calculated a Coefficient of Variation (CV) on the monthly ride counts:
- Low CV -> reliable commuter demand
- High CV -> seasonal, weather-sensitive usage
Casual riders show ~5% churn increase starting in November, while members continue riding despite seasonal changes.
Importantly, casual retention remains high suggesting many unsubscribed are residents, not only tourists.
I bucketed ride duration and analyzed return rates within short time windows to evaluate onboarding success.
The 3–15 minute range emerges as the system’s sweet spot: long enough to experience value, short enough to feel efficient.
Short rides (<3 minutes) were classified as:
-
Technical failure: short + same-station return
-
Last-mile success: short + different-station return
Findings:
Last-mile success:
- Members: 298,854 rides
- Casuals: 30,015 rides
Technical failures:
- Casuals: 12,837 rides
- Members: failures account for only ~9% of their short trips
For casual riders, roughly 1 in every 2.3 short trips fails due to technical issues. Members appear either better at identifying faulty bikes or more tolerant due to subscription costs.
To see if short round trips reflect failure or leisure exploration, I compared weekday vs. weekend behavior.
Hypothesis:
- Failures should occurr at similar rates on weekdays and weekends
- Exploration should spike on weekends
At key leisure stations, short round trips increase 22–24% on weekends, indicating that most of these rides are intentional exploration, not system failure.
I defined a gold standard member profile across:
- Time (when rides occur)
- Directionality (point-to-point vs loop)
- Efficiency (duration)
- Consistency (weekday vs. weekend balance)
Both members and casuals (~80%) prefer electric bikes, making hardware s secondary differentiator.
The Member Profile:
- Median ride is only 8.5 mins
- Nearly 78% of their rides happened during the work week and almost half (47.7%) during rush hour
- Low round-trip rate of 1.7%
- High-frequency, short-duration, weekday-heavy, point-to-point utility.
The Casual Profile:
- Average duration roughly 2x longer
- They're less likely to ride during rush hour (39%)
- 3.5x higher round-trip rate (6.1%)
- Low-frequency, long-duration, weekend-leaning, exploratory loops.
Conclusion: There's a Subset of Casual riders who represent 64% of casual volume on weekdays and roughly 39% during rush hour. They act like members. The Churn isn't a problem in the product, the subscription conversion is.
In this graph above we can see the hourly dynamic of User's behavior based on their type (Member vs. Casual).
Goal: Identify structural imbalances that require manual intervention.
I analyzed net flow by station:
- Sources: net negative flow (bikes leave)
- Sinks: net positive flow (bikes accumulate)
Key Finding: Transit hubs act as morning sources but do not reverse in the evening. Riders depend on bikes to get to work but switch modes to return home.
As a result:
- Transit hubs end the day in bike debt
- The system cannot self-correct without vans
Residential and park stations show artificial inflow spikes in early morning, likely caused by overnight rebalancing so commuters can deplete them by 9AM.
Without this intervention, the morning commute would fail.
Conclusions:
-
Stations that do not naturally rebalance to zero should be prioritized for maintenance and monitoring.
-
One-way hubs should receive bike drop-offs ~30 minutes before depletion begins, reducing van idle time and fuel costs.
Since 79% of all rides on electric bikes Electric, battery logistics become a critical operational constraint. Classic bikes can serve as a "risk buffer" as they:
- don't have charging dependency
- provide fallback capacity
Goal: Study how different stations operate in case of scaling to another city.
Stations were classified into:
- Commuter Hubs: rush-hour dominant
- Leisure Zones: weekend and late-evening dominant
- Hybrid Zones: balanced usage
1. Commuter Hubs
- Highest volume (15–16K average/median rides)
- No strong preference between classic and electric
- Most stable demand profile
2. Hybrid Zones
- Largest gap between classic and electric usage
- Members strongly prefer electric bikes
3. Leisure Zones
- Highest electric + member usage (~17K average)
- Even leisure riding is dominated by subscribed users
1. Commuter Hubs:
- Optimize dock availability and charge readiness
- Stability over experimentation
2. Hybrid Zones
- Replace older classic bikes with electric to unlock incremental revenue
3. Leisure Zones
- Stage newest, highest-quality e-bikes before weekends
- Members are power users even in recreational contexts
After end-to-end analysis I came to the conclusion that the biggest challenge is aligning user intent with pricing, hardware and operations.
What can be done in the future:
- convert existing behavior into subscriptions
- reduce operational drag
- scale to new cities with a repeatable playbook