Which Divvy bike station needs another rack?
Business Problem Overview
Problem statement
How can we determine which Divvy bike station would benefit from another rack?
Hypothesis / assumption to test
There exists at least one Chicago station where adding a rack would relieve user pain or cut Divvy costs.
Initial hunch: docks near high-traffic leisure spots (lakefront, bar clusters) will top the list.
Divvy strategy to keep in mind
City + Lyft are still in “expand-everywhere” mode: 250 new stations and thousands of e-bikes are being rolled out by 2025, with an explicit focus on underserved South- and West-side neighborhoods and on reducing re-balancing van mileage. (chicago.gov, divvybikes.com)
Benefits to track
- Customer pain ⇒ difficulty docking when a station is full.
- Divvy profit ⇒ cost of truck runs to rebalance bikes.
Success metrics
| Stake-holder | Metric we can calculate | Why it works |
|---|---|---|
| Riders | % Time Full (station effectively full) | Directly measures “can’t dock” pain |
| Divvy Ops | Net Accumulation per Dock = (Ends − Starts) ⁄ Dock Count | Positive number ⇒ trucks must clear bikes |
Format for our answer
-
Dock-Pressure Index (DPI) – a single score
DPI = (Ends − Starts) / Dock_Count × %Time_Full -
Either
- the station with the highest DPI, or
- a threshold list of stations whose DPI exceeds that cutoff.
Higher DPI → stronger business case for more racks.
Data Findings
Results
- Initial reaction: Two stations (West Loop and Chinatown) show the top DPI by a ~50% margin
- Interesting findings
- West- & South-side docks rank highest; likely due to fewer re-balancing visits (this is against my initial hunch)
- Event-centric docks (Lakefront path, Southport shopping, Wrigley Field) show the **highest overflow (**Divvy is popular for one-off trips to events).
- Two Wells St stations by a strip of bars show **lowest overflow (**people ride home drunk but start elsewhere)
- 56% of stations were never full in the two weeks before 6/3/25.
- A handful sit > 90% full but have near-zero overflow – high churn but balanced flows.
- System median overflow is just 0.07, so overall placement is solid of stations by Divvy.
Synthesis
The handful of high-DPI docks represent structural pain points inside an otherwise balanced network (median overflow is .07 and most stations are never full). They create both rider dissatisfaction and avoidable van trips – prime targets for extra racks.
Recommendation – what Divvy should do now
| Priority | Action | Impact |
|---|---|---|
| High | Install 1-2 racks at the top-DPI station(s). | Immediate customer relief + reduced van miles. |
| Medium | Re-route balancing vans toward high-DPI quadrants, away from low-utilization docks. | Cuts fuel and labor costs. |
| Low | Monitor post-install DPI to validate improvement; iterate threshold. | Confirms ROI. |
Looking ahead – where Divvy should focus next
- Peak-weighted % Time Full – weight the metric by demand hours to filter out overnight false positives.
- Capture “failed dock” attempts in the app to move from proxy pain to actual pain.
- Revenue lens – model upside from longer ride minutes or membership retention, not just cost savings.
- E-bike stray distance – another proxy for docking friction, especially where charging docks are scarce. (planetizen.com)
Main Critique (MOO) of the current approach
- % Time Full isn’t demand-weighted. A dock that’s full at 3 a.m. shouldn’t count the same as 5 p.m. rush.
- We only see successful docks. Attempted/diverted docks are invisible, so pain is understated.
- Focuses on cost avoidance, not revenue growth. Divvy may already be tracking re-balancing costs; the model should also test revenue-positive scenarios.
What I’d change next time
- Build a peak-hour weighting for % Time Full.
- Instrument the app to log failed dock attempts for ground-truth pain.
- Attach dollar values (lost rides, truck cost per mile) to each DPI point for a clearer go / no-go threshold.
- Get more creative in how to determine the benefit metrics for reducing user pain and increasing Divvy profits.