Cyclistic Bike-Share Analysis

Introduction

Cyclistic has grown as a leading bike-share program in Chicago, serving thousands of users with diverse needs through its network of more than 5,800 bicycles and 600 docking stations. To continue its growth, Cyclistic aims to convert casual riders into annual members. This report provides an in-depth analysis of how annual members and casual riders use Cyclistic bikes differently and offers data-driven recommendations to maximize membership conversions.

This case study follows the data analysis process: Ask, Prepare, Process, Analyze, Share, and Act.

Business Task

The primary business objective is to answer the following question: How do annual members and casual riders use Cyclistic bikes differently? This analysis will guide marketing strategies to convert casual riders into annual members by identifying behavioral patterns and actionable insights.

Key Stakeholders:

• Lily Moreno: Director of Marketing, responsible for campaign development.

• Cyclistic Marketing Analytics Team: Analysts (including you) collecting and analyzing data to support the marketing strategy.

• Cyclistic Executive Team: Decision-makers who will approve the recommended marketing strategy based on data insights.

Desired Outcome:

• Identify behavioral patterns between casual riders and annual members.

• Generate actionable insights supported by data visualizations.

• Provide strategic recommendations to increase membership conversions.

Data Sources and Preparation for Cyclistic Bike-Share Analysis

Data Source Overview:
The data used for this analysis comes from Cyclistic's publicly available historical trip datasets, provided by Motivate International Inc. under a public data license. These datasets represent actual bike-share usage patterns over the past 12 months. The data can be accessed and downloaded from the Cyclistic data repository provided in the case study.

Data Structure and Organization:

• File Format: The data is available as CSV files. Each file represents a month's worth of trip data, making up a full year's data when combined.

• Data Granularity: Each row in the dataset represents a single bike trip. The columns provide details such as:

  • ride_id: Unique identifier for each trip.

  • rideable_type: Type of bike used (standard, electric, or assistive).

  • started_at & ended_at: Timestamps for trip start and end.

  • start_station_name & end_station_name: Station details.

  • member_casual: User type (either member for annual subscribers or casual for single-use and day-pass riders).

Data Integrity and Quality Check (ROCCC Framework):
To ensure the data is reliable for our analysis, we evaluate it against the ROCCC framework:

• Reliability: The data originates from the official Cyclistic system managed by Motivate International Inc., a reputable bike-share operator.

• Originality: The data is directly sourced from Cyclistic's bike system, ensuring first-hand accuracy.

• Comprehensive: The dataset includes essential variables for behavior analysis, such as ride duration, start and end locations, and user type.

• Current: The dataset covers the most recent 12 months, making it relevant for current trends and decision-making.

• Cited: The data source is clearly cited and publicly accessible under a public license.

Potential Data Quality Issues:

• Missing Values: Some rows may have missing station names or coordinates.

• Inconsistent Labels: The member_casual column may have legacy labels (Subscriber instead of member).

• Outliers: Negative or zero ride durations may appear due to system tests or data errors.

• Data Cleaning Requirement:

  • Convert data types where needed.

  • Standardize member_casual values.

  • Remove rows with negative ride durations.

Data Validation Steps Taken:

• Inspected the first few rows using head() to ensure structural consistency.

• Checked for missing values using sum(is.na()).

• Verified data types with str().

• Confirmed integrity by comparing the total rows across all monthly datasets.

Data Relevance for Business Task:
This dataset provides the necessary information to analyze user behavior patterns for both casual and annual members. It enables us to:

• Identify ride duration patterns.

• Examine usage frequency and timing.

• Compare bike usage trends across membership types.

Data Processing and Cleaning Documentation for Cyclistic Bike-Share Analysis

Objective:
This section outlines the steps taken to prepare and clean the Cyclistic bike-share dataset, ensuring data quality and readiness for analysis. The goal is to process the data effectively to identify trends in how annual members and casual riders use the service differently.

Tools Used and Rationale:

• R Programming Language: Chosen for its powerful data manipulation packages and statistical capabilities.

• Packages Used:

  • tidyverse: For data wrangling and visualization.

  • conflicted: To handle conflicts between functions from different packages.

  • dplyr: Core part of tidyverse used for data transformation.

Step 1: Data Collection

Action:

• Imported quarterly datasets using read_csv().

• Datasets used: Divvy_Trips_2019_Q1.csv and Divvy_Trips_2020_Q1.csv.

Code Used:

R

Copy code

q1_2019 <- read_csv("Divvy_Trips_2019_Q1.csv")

q1_2020 <- read_csv("Divvy_Trips_2020_Q1.csv")

Step 2: Data Wrangling and Merging Datasets

Actions Taken:

• Checked Column Consistency: colnames() was used to compare column names for consistency across datasets.

• Renamed Columns: Legacy column names were renamed to match the latest schema.

• Converted Data Types: The ride_id and rideable_type columns were converted to character for consistent data types across datasets.

• Merged Dataframes: The datasets were combined using bind_rows().

Code Used:

R

Copy code

colnames(q1_2019)

colnames(q1_2020)

q1_2019 <- rename(q1_2019,

                  ride_id = trip_id,

                  rideable_type = bikeid,

                  started_at = start_time,

                  ended_at = end_time,

                  start_station_name = from_station_name,

                  start_station_id = from_station_id,

                  end_station_name = to_station_name,

                  end_station_id = to_station_id,

                  member_casual = usertype)

q1_2019 <- mutate(q1_2019, 

                  ride_id = as.character(ride_id), 

                  rideable_type = as.character(rideable_type))

all_trips <- bind_rows(q1_2019, q1_2020)

Step 3: Data Cleaning and Transformation

Actions Taken:

1. Removed Unnecessary Columns:

   • Columns dropped due to irrelevance and inconsistencies (start_lat, start_lng, end_lat, end_lng, birthyear, gender, and tripduration).

all_trips <- all_trips %>%

   select(-c(start_lat, start_lng, end_lat, end_lng, birthyear, gender, "tripduration"))

2. Standardized Membership Labels:

   • The member_casual column contained inconsistent labels (Subscriber and Customer).

   • Replaced with the consistent labels: "member" and "casual".

all_trips <- all_trips %>%

    mutate(member_casual = recode(member_casual, 

                                  "Subscriber" = "member", 

                                  "Customer" = "casual"))

3. Generated Date-Based Columns for Aggregation:

   • Extracted the day, month, and year to enable time-based analysis.

all_trips$date <- as.Date(all_trips$started_at)

all_trips$month <- format(as.Date(all_trips$date), "%m")

all_trips$day <- format(as.Date(all_trips$date), "%d")

all_trips$year <- format(as.Date(all_trips$date), "%Y")

all_trips$day_of_week <- format(as.Date(all_trips$date), "%A")

4. Calculated Ride Duration (ride_length):

   • Ride duration was computed as the difference between ended_at and started_at.

all_trips$ride_length <- difftime(all_trips$ended_at, all_trips$started_at)

all_trips$ride_length <- as.numeric(as.character(all_trips$ride_length))

Step 4: Error Handling and Data Integrity Checks

Actions Taken:

• Negative Values Removal:

  • Rides with negative durations were removed, as they indicated test rides or quality control.

• Removed Quality Control Entries:

  • Removed rows where start_station_name == "HQ QR" which was identified as a test station.

all_trips_v2 <- all_trips[!(all_trips$start_station_name == "HQ QR" | all_trips$ride_length < 0),]

Data Quality Verification:

• Checked Column Consistency: colnames() and str()

NA Handling: NA values were checked but not explicitly mentioned in the code. Consider using:
sum(is.na(all_trips))

Step 5: Data Integrity Verification

✅ Verification Actions Taken:

• Column Names Verified: colnames(all_trips)

• Data Types Confirmed: str(all_trips)

Corrected Factor Levels for Days of the Week:

all_trips_v2$day_of_week <- ordered(all_trips_v2$day_of_week,                                 levels=c("Sunday","Monday","Tuesday","Wednesday", "Thursday" "Friday", "Saturday"))

4. Data Exploration and Analysis Summary

4.1 Overview

This analysis aims to explore the key behavioral differences between annual members and casual riders using Cyclistic bikes. By examining ride frequency, duration, timing, and station usage patterns, we uncover insights into how these two groups use Cyclistic services. These findings directly address the business question: How do annual members and casual riders use Cyclistic bikes differently?

4.2 Descriptive Statistics

 To understand central tendencies, variability, and outliers in ride length for both user types.

Code used: 

mean(all_trips_v2$ride_length)  

median(all_trips_v2$ride_length)  

max(all_trips_v2$ride_length)  

min(all_trips_v2$ride_length)  

summary(all_trips_v2$ride_length)  

mean(all_trips_v2$ride_length)

[1] 1189.459

> median(all_trips_v2$ride_length)

[1] 539

> max(all_trips_v2$ride_length)

[1] 10632022

> min(all_trips_v2$ride_length)

[1] 1

> summary(all_trips_v2$ride_length)

    Min.  1st Qu.   Median     Mean  3rd Qu.     Max. 

       1      331      539     1189      912 10632022

Key Findings:

• Outliers: Extreme values were identified using max() and min() functions, suggesting potential non-standard rides.

• Data Consistency: No missing values were identified during summary inspections.

4.3 User Type Comparison

Objective:
Compare how casual riders and annual members use Cyclistic bikes differently by aggregating ride duration data.

Actions Taken:

• Calculated the mean, median, maximum, and minimum ride_length for each user group.

• Grouped data by both user type (member_casual) and day of the week.

Code Used:

# Summary stats by user type

aggregate(all_trips_v2$ride_length ~ all_trips_v2$member_casual, FUN = mean)

member_casual ride_length

1        casual   5372.7839

2        member    795.2523

aggregate(all_trips_v2$ride_length ~ all_trips_v2$member_casual, FUN = median)

member_casual ride_length

1        casual        1393

2        member         508

Key Insights:

• Casual Riders: Longer average ride durations, indicating more recreational use.

• Annual Members: Shorter, more consistent ride durations, suggesting frequent but shorter trips, likely for commuting.

4.4 Weekly Usage Patterns

To Identify how ride behavior varies by the day of the week for both casual riders and annual members.

Aggregated the data to calculate the average ride duration for each combination of member_casual and day_of_week.

Code used: 

aggregate(all_trips_v2$ride_length ~ all_trips_v2$member_casual + all_trips_v2$day_of_week, FUN = mean)

Output- Table:

member_casual day_of_week ride_length

1         casual      Sunday   5061.3044

2         member      Sunday    972.9383

3         casual      Monday   4752.0504

4         member      Monday    822.3112

5         casual     Tuesday   4561.8039

6         member     Tuesday    769.4416

7         casual   Wednesday   4480.3724

8         member   Wednesday    711.9838

9         casual    Thursday   8451.6669

10        member    Thursday    707.2093

11        casual      Friday   6090.7373

12        member      Friday    796.7338

13        casual    Saturday   4950.7708

14        member    Saturday    974.0730

Key Insights:

• Casual Riders: High usage on weekends, indicating recreational use.

• Annual Members: Consistent usage throughout the workweek, aligning with commuter behavior.

4.5 Ridership Trends Over Time

Analyze ride frequency and average duration trends across the week to assess usage patterns more deeply.

Actions Taken:

• Created a grouped summary using dplyr.

• Visualized the distribution using ggplot2.

Code used: 

all_trips_v2 %>%

    mutate(weekday = wday(started_at, label = TRUE)) %>%

    group_by(member_casual, weekday) %>%

    summarise(number_of_rides = n(), 

              average_duration = mean(ride_length)) %>%

    arrange(member_casual, weekday)

Output Table: 

# A tibble: 14 × 4

# Groups:   member_casual [2]

   member_casual weekday number_of_rides average_duration

   <chr>         <ord>             <int>            <dbl>

 1 casual        Sun               18652            5061.

 2 casual        Mon                5591            4752.

 3 casual        Tue                7311            4562.

 4 casual        Wed                7690            4480.

 5 casual        Thu                7147            8452.

 6 casual        Fri                8013            6091.

 7 casual        Sat               13473            4951.

 8 member        Sun               60197             973.

 9 member        Mon              110430             822.

10 member        Tue              127974             769.

11 member        Wed              121902             712.

12 member        Thu              125228             707.

13 member        Fri              115168             797.

14 member        Sat               59413             974.

>

4.5.1 Visualization- Number of Rides by Rider Type

Created a bar plot comparing casual and member usage by weekday.

Code Used:

all_trips_v2 %>%

    mutate(weekday = wday(started_at, label = TRUE)) %>%

    group_by(member_casual, weekday) %>%

    summarise(number_of_rides = n(), 

              average_duration = mean(ride_length)) %>%

    ggplot(aes(x = weekday, y = number_of_rides, fill = member_casual)) +

    geom_col(position = "dodge") +

    labs(title = "Number of Rides by Rider Type", 

         x = "Day of the Week", 

         y = "Number of Rides") +

    theme_minimal()

The bar plot displays the total number of rides for casual riders and annual members across the days of the week. Each bar is split by rider type, with distinct colors representing casual and member riders.

Key Insights

1. Casual Riders:

   • Exhibit a significant preference for weekends (Saturday and Sunday), with ride numbers peaking on these days.

   • Weekday usage is relatively low, indicating that casual riders may primarily use the service for leisure or recreational purposes.

2. Annual Members:

   • Maintain a consistent pattern of usage across all weekdays, with slightly higher activity from Monday to Friday.

   • This indicates that annual members might rely on Cyclistic bikes for daily commuting or routine travel needs.

• Strategic Implications:

  • Marketing campaigns aimed at converting casual riders into members should emphasize the benefits of weekday rides, such as convenience for daily commuting.

  • Promotions or incentives can target casual riders during weekends to encourage frequent use, potentially transitioning them into memberships.

  • Highlighting cost savings for frequent riders and accessibility features for weekday commutes could resonate with casual riders.

4.5.2 Analysis: Average Ride Duration by Rider Type

Objective:
The goal of this analysis is to compare the average duration of bike rides for casual riders and annual members across the days of the week.

Code Used:

all_trips_v2 %>%

    mutate(weekday = wday(started_at, label = TRUE)) %>%

    group_by(member_casual, weekday) %>%

    summarise(number_of_rides = n(), 

              average_duration = mean(ride_length)) %>%

    ggplot(aes(x = weekday, y = average_duration, fill = member_casual)) +

    geom_col(position = "dodge") +

    labs(title = "Average Ride Duration by Rider Type", 

         x = "Day of the Week", 

         y = "Average Duration (Seconds)") +

    theme_minimal()

Key Insights :

1. Significant Difference in Ride Duration:

   1. Casual riders consistently have much longer average ride durations compared to members across all days of the week.

   2. This suggests that casual riders may use the bikes more for leisure or exploratory purposes, taking extended rides.

2. Consistency for Members:

Annual members show stable and significantly shorter average ride durations throughout the week, averaging around 500-600 seconds (8-10 minutes).

This consistency indicates that members likely use the service for commuting or other short, practical trips.

4.6 Ride Frequency Analysis (Usage Patterns Over Time)

Objective:

Identify the frequency of rides across time intervals (e.g., hourly, daily, monthly trends).

Analysis Approach:

• Time of Day Analysis:

  • Aggregate the number of rides taken by hour of the day (hour extracted from started_at).

Code used:

all_trips_v2 %>%

    mutate(hour = lubridate::hour(started_at)) %>%

    group_by(member_casual, hour) %>%

    summarise(number_of_rides = n()) %>%

    ggplot(aes(x = hour, y = number_of_rides, color = member_casual)) +

    geom_line(size = 1.2) +

    labs(title = "Ride Frequency by Hour", x = "Hour of the Day", y = "Number of Rides")

4.6.1 Visualization: Line Chart

4.6.2 Heatmap of Rides (Time vs. Day)

Objective:

Understand peak riding hours for both groups.

Analysis Approach:

• Use a heatmap showing the count of rides by hour and day of the week.

Code Example:

all_trips_v2 %>%

    mutate(hour = lubridate::hour(started_at)) %>%

    group_by(day_of_week, hour, member_casual) %>%

    summarise(number_of_rides = n()) %>%

    ggplot(aes(x = hour, y = day_of_week, fill = number_of_rides)) +

    geom_tile() +

    facet_wrap(~ member_casual) +

    labs(title = "Heatmap of Rides by Hour and Day", fill = "Number of Rides")

Key Findings:

1. Annual Members:

   • Primarily use bikes for weekday commuting, with activity peaking during morning (7–9 AM) and evening rush hours (4–6 PM).

   • Their usage is concentrated on Monday through Friday, with minimal activity on weekends.

2. Casual Riders:

   • Ride for leisure or recreational purposes, with higher activity during weekends.

   • Their usage peaks during midday to early evening hours (10 AM–6 PM), showing more relaxed and flexible timing.

4.7 Station Usage Analysis (Popular Stations and Routes)

Objective:

Identify the most frequently used starting stations for both annual members and casual riders to understand their preferred locations and plan targeted strategies.

• Count the frequency of trips from each station.

• Identify the top 10 stations for casual riders and annual members.

Code Example:

all_trips_v2 %>%

    group_by(member_casual, start_station_name) %>%

    summarise(number_of_rides = n()) %>%

    arrange(desc(number_of_rides)) %>%

    slice(1:10)

# A tibble: 20 × 3

# Groups:   member_casual [2]

   member_casual start_station_name           number_of_rides

   <chr>         <chr>                                  <int>

 1 casual        Streeter Dr & Grand Ave                 2749

 2 casual        Lake Shore Dr & Monroe St               2732

 3 casual        Shedd Aquarium                          1832

 4 casual        Millennium Park                         1406

 5 casual        Michigan Ave & Oak St                   1017

 6 casual        Michigan Ave & Washington St             839

 7 casual        Dusable Harbor                           832

 8 casual        Adler Planetarium                        827

 9 casual        Theater on the Lake                      795

10 casual        Lake Shore Dr & North Blvd               606

11 member        Canal St & Adams St                    13799

12 member        Clinton St & Washington Blvd           13434

13 member        Clinton St & Madison St                12891

14 member        Kingsbury St & Kinzie St                8720

15 member        Columbus Dr & Randolph St               8515

16 member        Canal St & Madison St                   7957

17 member        Franklin St & Monroe St                 7010

18 member        Michigan Ave & Washington St            6686

19 member        Larrabee St & Kingsbury St              6467

20 member        Clinton St & Lake St                    6439

Strategic Implications:

• For Casual Riders:
  The high concentration of rides starting near tourist and leisure spots highlights the opportunity for marketing campaigns promoting memberships at these locations. Targeted offers or advertisements placed at these stations could encourage conversion.

• For Members:
  The data validates that members use bikes as a reliable part of their daily commute. Cyclistic could explore partnership opportunities with businesses in these areas to increase visibility and value for members.

Strategic Implications:

• Cyclistic might want to examine whether stations with high casual rider usage experience capacity issues during peak times and allocate resources accordingly.

• Implementing digital campaigns or promotions (e.g., discounted annual memberships) specifically at popular casual rider stations could influence conversions.

• Compare the busiest times for these stations by user type to optimize station operations and scheduling.

Share

Top 3 Recommendations

1. Target Casual Riders with Weekend-to-Weekday Transition Incentives

   • Insight from Analysis: Casual riders primarily use Cyclistic bikes on weekends for leisure, while annual members use bikes consistently throughout the week, likely for commuting.

   • Recommendation: Launch a marketing campaign that incentivizes casual riders to use bikes during weekdays. For example, offer discounted weekday rides or a "Weekday Challenge" where casual riders earn rewards for using bikes on weekdays. Highlight the convenience and cost savings of annual memberships for frequent weekday use.

   • Strategic Impact: This could encourage casual riders to adopt biking as a regular commuting option, increasing the likelihood of converting them to annual members.

2. Enhance Station Capacity and Visibility at Popular Casual Rider Locations

   • Insight from Analysis: Casual riders tend to start their rides at popular tourist and leisure spots (e.g., Streeter Dr & Grand Ave, Lake Shore Dr & Monroe St).

   • Recommendation: Increase bike availability and docking capacity at these high-traffic stations, especially during peak weekend hours. Additionally, place targeted advertisements or promotional materials at these stations to promote annual memberships.

   • Strategic Impact: Improved station capacity will enhance user experience, while targeted promotions can directly influence casual riders to consider annual memberships.

3. Leverage Digital Media to Highlight Cost Savings and Convenience for Commuters

   • Insight from Analysis: Annual members have shorter, more consistent ride durations, suggesting they use bikes for commuting. Casual riders, on the other hand, take longer rides, primarily on weekends.

   • Recommendation: Use digital media (social media, email campaigns, and in-app notifications) to emphasize the cost savings and convenience of annual memberships for daily commuters. Create content that compares the cost of casual rides versus annual memberships, showcasing the financial benefits of membership.

   • Strategic Impact: This approach can appeal to casual riders who may not realize the long-term savings of an annual membership, especially if they start using bikes more frequently during weekdays.

Conclusion

The analysis clearly differentiates the behavioral patterns of annual members and casual riders. These findings provide actionable insights to guide Cyclistic’s marketing and operational strategies. By focusing on tailored promotions, resource optimization, and commute-centric enhancements, Cyclistic can achieve its goal of increasing membership conversions and enhancing customer satisfaction.