🚲 Cyclistic Bike Share Analysis How do annual members and casual riders use Cyclistic bikes differently?
Portfolio Project | Business Analyst / Data Analyst Tools: Python · SQL · Power BI / Tableau · Excel Dataset: Divvy Trip Data — April 2024 (415K rides)
📌 Business Context Cyclistic is a Chicago-based bike-share company with 5,800+ bicycles and 600+ docking stations. The marketing team believes that converting casual riders (single-ride or day-pass users) into annual members is key to long-term growth. The core question:
What behavioural differences exist between casual riders and annual members — and how can these insights drive a targeted marketing strategy?
📁 Project Structure cyclistic-analysis/ │ ├── 📓 cyclistic_analysis.py # Python EDA & visualizations ├── 🗄️ cyclistic_sql_analysis.sql # SQL business queries (10 questions) ├── 📊 cyclistic_charts.png # Output charts ├── 📋 cyclistic_summary.csv # Aggregated data (for Power BI / Tableau) ├── 🧹 cyclistic_clean.csv # Cleaned dataset └── 📝 README.md
🔍 Methodology
Source: Kaggle — Divvy Bike Sharing (April 2024) Raw data: 415,025 ride records, 13 columns Key fields: ride_id, rideable_type, started_at, ended_at, start/end station, lat/lng, member_casual
📊 Key Findings Finding 1 — Rider Split Rider TypeRidesShareAnnual Member282,98368.3%Casual131,33731.7% Finding 2 — Ride Duration Rider TypeAvg DurationMedianCasual21.4 min11.9 minMember11.8 min8.3 min
💡 Casual riders average nearly 2× longer per ride — indicating leisure/recreational use rather than commuting.
Finding 3 — Day of Week Pattern
Members ride heavily Mon–Fri (commute pattern), with a dip on weekends Casual riders peak on Saturday–Sunday (43% of casual rides are on weekends vs only 25% for members)
Finding 4 — Time of Day
Members: Clear twin peaks at 8AM and 5PM → classic commuter signature Casual riders: Single ramp-up peaking at 2–5PM → leisure pattern
Finding 5 — Top Casual Stations (Marketing Hotspots) StationCasual RidesCasual %Streeter Dr & Grand Ave3,31976%DuSable Lake Shore Dr & Monroe St2,48278%Michigan Ave & Oak St1,31956%
💡 Top casual stations are all in tourist/lakefront areas — not residential or office corridors.
💡 Business Recommendations Rec 1 — Weekend Membership Campaign
43% of casual rides happen on weekends
Launch a targeted weekend campaign: offer a discounted first month of annual membership to casual riders who ride 3+ times in a month. Activate via push notification right before their Saturday ride. Rec 2 — Afternoon In-App Nudge
Casual ride peak: 2–5PM
At 1:30PM on weekends, send a personalised cost comparison: “You’ve spent $X on day passes this month. An annual membership would have saved you $Y.” Rec 3 — Long-Ride Loyalty Program
Casual classic bike avg: 41 min/ride
Introduce a “Ride More, Pay Less” tier: after 5 rides exceeding 20 minutes, trigger a one-click membership upgrade offer at a 20% discount. Rec 4 — Station-Level Physical Marketing
Top 3 casual stations are all tourist/leisure corridors
Install membership sign-up QR codes at Streeter Dr, DuSable Lake Shore Dr, and Michigan Ave docks. These stations have 70–80% casual rider rates — highest conversion potential.
🛠️ Tools Used ToolPurposePython (pandas, matplotlib, seaborn)EDA, data cleaning, visualisationSQL (PostgreSQL-compatible)Business queries, aggregationsPower BI / TableauInteractive dashboardExcelExecutive summary, pivot tables
📈 Dashboard (Power BI / Tableau) The interactive dashboard covers 4 views:
Overview — KPIs, rider split, total rides Behaviour — Hourly & daily patterns by rider type Geography — Station-level heatmap (casual vs member concentration) Insights — Duration distribution, bike preference, conversion funnel
→ View Dashboard (link to be added)
📊 Live Dashboard
🔗 View Interactive Dashboard: https://hoanggialac.github.io/Cyclistric-bike-share-analysis/
pip install pandas numpy matplotlib seaborn
python cyclistic_analysis.py
📚 Data Source
Dataset: Divvy 2024–2025 Bike Sharing Data via Kaggle Original provider: Motivate International Inc. / Lyft Bikes and Scooters License: Divvy Data License Agreement Data has been anonymised — no personally identifiable information is included.