Background

Cyclistic, a Chicago-based bike share company wants to maximize the number of subscribers since they are more profitable than casual riders.

In order to do that, Cyclistic needs to be er understand how annual members (subscribers) and casual riders differ, why casual riders would buy a membership, and how digital media could affect their marketing tactics.

This report answers the first question:
- How do annual members and casual riders use Cyclistic bikes differently?

Executive summary

Text

• Bullet 1
• Bullet 2
• Bullet 3

Slide with recommendations

Text

Brief description of the dataset and the analysis

The analysis is performed by examining ridership data from April 2023 through March 2024 collected from this source:

https://divvy-tripdata.s3.amazonaws.com/index.html

The dataset consist of 5,750,177 rows. 165,844 rows were deleted during cleaning, and the final sample size is 5,584,333 rows.

See appendix 1 for a more detailed description the data cleaning process.

Members account for 64 percent of all rides and casual users account for 36 percent.

I use normalized data in my analysis in the form of percentage distributions and ride length.

All analysis and visualizations were peformed in R.

Casual users ride more often in weekends and in the summer months while members ride more often on regular weekdays and in the fall/winter seasons.

• Bike sharing service is used more often in the summer season, likely due to higher temperatures.
• The higher proportion of member users on regular weekdays and in the fall and winter seasons compared to casual users could indicate a higher proportion of commuters.
• The higher proportion of casual users in weekends and in the summer season could indicate a higher proportion non-commuters such as tourists and visitors.

Average ride length for casual riders are higher compared to members. Average ride length is higher in weekends and in the summer season for all users.

• Average ride length follow a season pattern where it increases when the weather is warmer. Warmer weather usually aligns with major holidays which could also indicate a higher proportion of non-commute users in these periods.
• The higher average ride length for casual users could indicate a higher proportion of non-commuters who are often have a more relaxed schedule

A higher proportion of casual users ride electric and docked bikes.

• The higher proportion of electric bike use among casual riders could indicate they travel longer distances. This could also explain the higher average ride length among casual users.

Choice of ride depends on the season. Casual users prefer electric bikes in the winter season. Members choice of bike have less variation.

• Members’ choice of bike may seem odd since they favor classic bikes in January and February when the weather is usually cold. Could be explained by specific weather conditions and/or holidays.
• Casual users’ choice of bike follow a more logical pattern where use of classic bikes increase in months where the temperature is higher.

Appendix 1: Data cleaning

R was used to clean data and look for errors.

The data cleaning uncovered some inconsistencies:

• 72 observations in the “started_at” variable could not be successfully converted to POSIXct date-time format.
• 85 observations in the “ended_at” variable could not be successfully converted to POSIXct date-time format.
• 321 observations contained inconsistent start/end ride data in the form of negative ride length.
• 7,566 observations contained missing GPS location data.
• Some outliers in ride length. The majority (97 percent) of the observations in the dataset have a ride length less than 3 hours. Some observations have a ride length less than 60 seconds and a few even 0 seconds (!) which seems highly unlikely. These outliers could skew the results.

165,844 (2.9 percent) “bad” observations and outliers described in the paragraphs above were removed from the dataset.

The final sample size is 5,584,333.

Appendix 2: Ride length and membership status

• Average ride length is higher (19 minutes) for casual users than members (12 minutes), and has a higher variance.
• The majority of rides took less than 10 minutes.

Note: Observations with a ride length below 1 minute and above 180 minutes were removed from the dataset . See appendix 1.

Appendix 3: Sample distribution by member status and month

Table 1: Ridership by member status and month:

Appendix 4: Ridership by member status and day of week

Table 2: Ridership by member status and day of week: