Divvy Bike: Analysis
Veti Comesongsri
2025-11-30
This is my first Data Analysis Project. I did it as part of Google Data Analytics course. There are three documents, Prepare and Process Data, Analysis, and Findings and Recommendations.
Question
“How do annual members and casual riders use Divvy bikes differently?”
Introduction
“Divvy is the bicycle sharing system in the Chicago metropolitan
area, currently serving the cities of Chicago and Evanston. The system
is owned by the Chicago Department of Transportation and has been
operated by Lyft since 2019.”
Source: Wikipedia
To answer the question “How do annual members and casual riders use Divvy bikes differently?”, the following factors will be analyzed to find the differences.
- Number of Rides
- Bike Type
- Time of Day
- Day of Week
- Month and Seasons
- Riding Duration
- Location
Some of these factors will be analyze together to find interaction between them.
After that, the results of the analysis will be discussed. Then the limitation of the results will also be discussed.
Number of Rides
Total Number of Rides by Rider Type
| Rider Type | No of Rides |
|---|---|
| member | 7,121,972 |
| casual | 3,962,010 |
| Total | 11,083,982 |
The number of rides by member riders was significantly higher than casual riders. The ratio of casual:member was about 1:1.80. This may create difficulty in interpreting some graph as number of rides will always be lower for casual group. However, the pattern of graph still be useful to see the differences.
Number of Rides Per Day
| Rider Type | min | max | mean | sd |
|---|---|---|---|---|
| casual | 52 | 18,235 | 5,419.99 | 4,197.93 |
| member | 462 | 19,045 | 9,742.78 | 4,433.86 |
The lowest ride per day for casual and member were January 14, 2024
and January 15, 2024, respectively. At that time Chicago had very cold
weather and dangerous wind chills due to polar
vortex.
The highest ride per day for casual riders was September 1, 2024. On that day, there were Bike the Drive event, last day of Chicago Jazz Festival, Chicago Cubs game, and several events. For members, the highest ride per day was on September 11, 2024.
Conclusion: Number of Rides
From the figures above, we can see that number of trips of casual riders were generally lower than member riders. During the cold months, casual riders consistently rode less than members. However, during warm months, the number of rides for casual group varied considerably and sometimes they rode more than member riders. This will be investigated further when seasons are analyzed.
Bicycle Type
There were three types of vehicles available for rent, classic bicycle, electric bicycle, and electric scooter.
| Bike Type | No of Rides |
|---|---|
| electric_bike | 6,137,575 |
| classic_bike | 4,808,871 |
| electric_scooter | 137,536 |
| Total | 11,083,982 |
The overall percentage of trips on each bike were classic bike 43.39%, electric bike 55.37%, and electric scooter 1.24%. The electric scooter percentage was unusually low.
The classic bike and electric bike had comparable usage in 2023 and 2024. However, in 2025, the usage of classic bike dropped while electric bike increased. There was a news that Divvy will phase out classic bike. Later, Divvy indicated that they will maintain the current number of classic bike and growth overall fleet with electric bike.
The chart also explains why the number of rides on electric scooters was very low. The scooter had records only from September 2024 to October 2024. The electric scooter data will be filtered out for analyses that involve bike type.
Bike Type by Rider Type
The casual and member riders had similar bike preferences. The casual rider rode more on electric bike (57%) than classic bike (43%). The member rider also rode more on electric bike (55%) than classic bike (45%). The ratios of bike type for both groups were comparable too.
Weekday and Weekend
The member riders had about the same bike preference on both weekday and weekend. The casual riders preferred electric bike a little bit more on weekdays.
Seasons and Bike Type
The member riders ride fewer electric bikes during cold seasons. This may be because the electric bike did not work well when weather was extremely cold. The casual riders maintained about the same bike preference for all seasons.
Conclusion: Bike Type
In general, both groups rode slightly more on electric bikes than on classic bikes. This preference held true when considering other factors like weekday/weekend and season. There was slightly change in percentage, but the differences between casual riders and member riders were not much to be considered.
Time of Day
Overall, riding started in the early morning and had a small peak around morning rush hour. Then, it had a higher peak around evening rush hour. Riding slowed down after rush hour end and down to lowest point around 3 am the next day.
For the casual riders, riding started in the early morning and had only one at the evening rush hour. For the members, riding had two peaks corresponding to morning and evening rush hours. This may indicate that member riders used Divvy bikes to commute to and from work.
To compare the pattern of riding throughout the day of two groups that have different sizes, density graph might be better to visualize the difference. The density graph normalized both groups to both have area under the graph equal to 100%.
The casual riders group started early in the morning and steadily increased to peak at evening rush hour and then declined. The member riders started early in the morning too but quickly increased to first peak at morning rush hour, then declined a bit and quickly increased to second higher peak around evening rush hour.
Time of Day and Weekday/Weekend
Within each rider group, the riding patterns on Sunday and Saturday were different from other days. Between groups, the riding patterns on weekdays were different and the same on weekends.
To see the pattern easier, the days of week were combined into weekend and weekday. On weekdays, the differences were the same but more pronounced. Also, casual riders had mini peak during morning rush hour on weekdays. It seems like some of them used Divvy bike to commute to work too but not as much as member riders.
On the weekend, both casual and member riders had almost the same pattern. Both groups did not have any sharp peak at any time on weekend.
The density plot showed that the patterns of weekend and weekday for both groups did not change across seasons. The cold seasons just had less people riding Divvy bikes.
Time of Day, Weekend/Weekday and Seasons
When considering seasons together with weekend/weekday, casual riders rode more than member riders on summer weekends and rode a lot less on winter weekends. For on weekdays, casual riders rode less than members in all seasons. They also rode a lot less on winter weekdays.
The density plot shows that the patterns of weekend and weekday for both groups did not change across seasons. The cold seasons just had less people riding Divvy bikes.
Conclusion: Time of Day
On weekend, the number of rides started to go up in the early morning and steadily went up to peak around 3 pm. Then, number of rides on the weekend went down until early morning of the next day and started the cycle again. On the weekday, the number of rides started early morning for both groups. The number of rides for the members rose sharply to peak at around 8 am then declined until 10 am. After 10 am the number of rides for members rose again and peaked at around 6 pm, then declined to lowest point in the early morning the next day. For casual riders, they had the same pattern on weekdays but with only very small peak around 8 am. After that the pattern was the same as member riders.
Day of Week
Overall, average number of rides per day was not very different on each day of the week. Sunday had the lowest average number, and Saturday had the highest average.
When separated by rider type, causal riders rode more on the days on weekends than on weekdays. Member riders reversed the pattern, they rode more on weekdays and less on weekends.
When taking seasons into account, the riding pattern held true. Casual riders rode more on weekends in every season except winter. During winter, the number of rides dropped a lot compared to other seasons. For member riders the riding pattern was held every season.
Conclusion: Day of Week
The casual riders rode more on weekends than weekday. The member riders rode more on weekdays. In general, member riders rode more than casual riders every day on every season except summer weekend.
Ride Duration
All Data
The ride duration is calculated from interval between started_at and ended_at.
| min | max | mean | sd | Total |
|---|---|---|---|---|
| 1 | 180 | 14.38 | 15.64 | 11,083,982 |
| Q0 | Q1 | Q2 | Q3 | Q4 | mode |
|---|---|---|---|---|---|
| 1 | 5.72 | 9.77 | 17.06 | 180 | 4.87 |
Please note that the data has been cleaned to remove ride duration less than 1 minute and ride duration longer than 180 minutes.
The distribution of riding time is very skewed to the right with very long tail. This meant majority of the rides were short, but there were a few that were very long.
| 0% | 10% | 20% | 30% | 40% | 50% | 60% | 70% | 80% | 90% | 100% |
|---|---|---|---|---|---|---|---|---|---|---|
| 1 | 3.6 | 5.02 | 6.43 | 7.98 | 9.77 | 12 | 15.02 | 19.69 | 29.01 | 180 |
From the density graph and percentiles table, 50% of rider rode longer than 10 minutes, 20% rode longer than 20 minutes, and 10% rode less than 29 minutes.
Average riding durations fluctuated from day to day. However, the trend line shows that riding time followed seasonal cycle. Riders used Divvy bikes during warm seasons longer than cold seasons.
Riding Duration and Rider Type
| Rider Type | min | max | mean | sd | Total |
|---|---|---|---|---|---|
| casual | 1 | 180.00 | 19.04 | 21.12 | 3,962,010 |
| member | 1 | 179.97 | 11.79 | 10.68 | 7,121,972 |
On average, casual riders (19.04 minutes) rode Divvy bike longer than member riders (11.79 minutes).
In general, casual riders rode longer than member riders. Average riding durations of casual riders fluctuated more than member riders. The trend line of casual riders also had larger seasonal variances than member riders.
| Rider Type | 0% | 10% | 20% | 30% | 40% | 50% | 60% | 70% | 80% | 90% | 100% |
|---|---|---|---|---|---|---|---|---|---|---|---|
| casual | 1 | 4.30 | 6.08 | 7.85 | 9.79 | 12.13 | 15.14 | 19.37 | 26.13 | 40.98 | 180.00 |
| member | 1 | 3.35 | 4.61 | 5.85 | 7.20 | 8.77 | 10.65 | 13.12 | 16.80 | 23.62 | 179.97 |
Both casual and member had right skew riding duration distribution. The
casual group had flatter and less skew distribution than member group.
This means more casual riders on longer trips than member riders. About
half of the member riders rode longer than 12 minutes compared to only
33% of member riders. About 20% of casual riders rode longer than 26
minutes compared to only 7% of member riders. Finally, about 10% of
casual riders rode longer than 41 minutes comapared to only 1% of member
riders.
Weekend and Weekday
On average, casual riders rode longer than member riders on both weekend and weekday. On weekends, casual riders rode 67% longer than member riders. On weekdays, casual riders rode 53% longer than member riders. Casual riders rode 24% longer on weekends than weekdays, while member riders rode only 14% longer on weekends than weekdays.
Seasons
Casual riders rode longer than member riders on every season. For example, casual riders rode 59% longer than member riders during summer, and they rode 46% longer during winter. Casual riders rode longer in warm seasons than cold seasons. They rode 24% longer in summer than in winter. The member riders rode more consistent, they rode 17% longer in summer than in winter.
Seasons and Weekend/Weekday
Weekend/weekday did not have much interaction with seasons. The seasonal patterns were the same for both weekend and weekday.
Bike Type
Divvy riders used classic bike longer than electric bike. Casual riders rode classic bikes 46% longer than electric bikes. The member riders rode classic bikes 12% longer than electric bikes. Casual riders rode classic bikes 54% longer than member riders, and they rode electric bikes 26% longer.
Weekday/Weekend and Bike Type
Weekend/weekday did not have much interaction with bike type. The bike
type patterns were the same for both weekend and weekday.
Season and Bike Type
Season also did not have much interaction with bike type. The bike type
patterns were the same every seasons.
Time of Day
Casual riders rode longer than member riders on every hour of the day. They started to ride longer and longer around 8 am until 10 am. After 2 pm, the ride duration started to decline until early morning of the next day, and the cycle started again. For member riders, the duration was not change much throughout the day.
Conclusion: Ride Duration
Casual riders rode longer than member riders even when consider riding duration with other factors - weekend/weekday, bike type, seasons and time of day. The other difference was that the duration of casual riders varied much more than member riders when considering other factors.
Location
The data was clean to filter out latitudes and longitudes that felt outside Divvy bike service area. In addition, to create unbiased data used in the location analysis, a random sample of 100,000 casual riders and 100,000 member riders was used. The smaller sample also helped ease the memory requirement of the computer used to run the analysis.
| member_casual | start_lat | start_lng | end_lat | end_lng | season | week_day | morning_afternoon | length_min | |
|---|---|---|---|---|---|---|---|---|---|
| casual:100000 | Min. :41.65 | Min. :-87.85 | Min. :41.65 | Min. :-87.85 | Spring:52117 | weekday:138910 | morning : 57907 | Min. : 1.000 | |
| member:100000 | 1st Qu.:41.88 | 1st Qu.:-87.66 | 1st Qu.:41.88 | 1st Qu.:-87.66 | Summer:85271 | weekend: 61090 | afternoon:142093 | 1st Qu.: 5.943 | |
| NA | Median :41.90 | Median :-87.64 | Median :41.90 | Median :-87.64 | Fall :44729 | NA | NA | Median : 10.217 | |
| NA | Mean :41.90 | Mean :-87.65 | Mean :41.90 | Mean :-87.65 | Winter:17883 | NA | NA | Mean : 15.374 | |
| NA | 3rd Qu.:41.93 | 3rd Qu.:-87.63 | 3rd Qu.:41.93 | 3rd Qu.:-87.63 | NA | NA | NA | 3rd Qu.: 18.141 | |
| NA | Max. :42.07 | Max. :-87.53 | Max. :42.07 | Max. :-87.53 | NA | NA | NA | Max. :179.955 |
Chicago Population Density
For reference, the Chicago population density is included to compare
with bikes usage density. The density data is from Forest Gregg’s github
“Chicago-dots”
Start Locations
Both casual riders and member riders usages were concentrated in the downtown Chicago area and extended to the north along Lake Michigan shoreline, which aslo happen to have high population density too. There was also a small high density area around University of Chicago.
The zoomed density map paint a clearer differences between casual rider and member riders. Casual riders started location main hotspot was Navy Pier. The others were Michigan Avenue from loop to Chicago River, John Hancock Building, Union Station, and Millennium Park. Member riders’ main hotspots were around Union Station and Ogilvie Transportation Center. The other hotspot was area between brown line and red line from loop to Near North Side.
In summary, casual riders’ starting locations were around tourist spots while member riders’ starting locations were around business areas and major transportation hubs.
End Location
The ended location hotspots for both casual and member riders were almost the same as started location.
Weekday
On weekdays, casual riders’ hot spots were Navy Pier, Michigan Avenue near Millennium Park, Union Station, and area between brown line and red line in the Near North Side. Member rider’s major hotspots were area around Union Station and Ogilvie Transportation Center. The other was area between brown line and red line from loop to Near North Side.
Overall, both groups had high usage in the central business district. However, casual riders had lower density than member riders in every area except near Navy Pier.
Weekend
Area around Navy Pier was a major hotspot for casual riders on weekends. The next high-density area was Michigan Avenue near Millennium Park. Other areas were Magnificent Mile, Michigan Avenue from Museum Campus to Chicago River and area between brown line and red line in Near North Side. The area of high usage extended along Lake Michigan up to Lake View.
For member riders, the main hot spot was in the area between brown line and red line in Near North Side. The area of high usage also extended along Lake Michigan up to Lake View.
Seasons
The casual Riders’ Navy Pier hotspot disappeared during cold seasons.
The member riders’ hotspot pattern held about the same in every
season.
On weekends, Navy Pier hotspot also became less prominent during cold
seasons. Member riders’ hotspot patterns held about the same on every
season.
University of Chicago and Illinois Institute of Technology
The areas around major universities deserve special attention, because it is posible for student to behave differently than other groups. For student, Divvy has a cheaper University Membership.
Both casual riders’ and member riders’ hotspots were almost the same for University of Chicago area. The exception was casual riders hotspots were include Margaret T Borroughs Beach and Griffin Museum of Science and Industry. This indicated that both casual student riders and student member riders had the same behavior. In addition, there were some non-student casual rider in the area too.
Conclusion: Location
Casual and member riders used Divvy bikes at different locations. This can also imply that they used it for different purposes. Casual riders used bikes around tourist attractions and shopping areas in both weekday and weekend. Casual riders also used Divvy bikes to travel up north along Lake Michigan shoreline. Member riders mostly used Divvy bikes near transportation hubs and in the business district during weekday and used it to in the more concentrate around shopping area during weekend. Member riders also use it to travel north along lake Michigan shoreline during weekend.
From the pattern of the usage locations, we may imply that majority of casual riders are visitors or tourists that used bikes for travel along tourist destinations, while most of the member users are people who use bikes to commute on weekday and for leisure on weekends.
Routes
Top Routes
The top routes for casual riders were the same on weekends and weekdays. For member riders, the top routes on weekends were different from on weekdays.
Please note that there were some locations where the Divvy bike users started and later returned the bikes at the same location.
Top 20 Routes, Casual Riders Weekday
Casual riders’ top 20 routes on weekday were from and to the same set of stations. All the stations are near tourist attractions. In addition, there were many routes that start and end at the same stations.
Top 20 Routes, Casual Riders Weekend
Casual riders’ top 20 routes on weekend were the same as weekday. They were started and ended from tourist attraction and there were many route that started and end at the same stations.
The casual riders’ top 20 routes for both weekday and weekend involved only 12 stations.
Top 20 Routes, Member Riders Weekday
Member riders’ top 20 routes on weekday were from and to the larger set of stations than casual riders’. All the stations are in Loop, Near North Side and University of Illinois Chicago. there were no route that start and end at the same stations.
The member riders’ top 20 routes for weekday involved 21 stations.
Top 20 Routes, Member Riders Weekday
Member riders’ top 20 routes on weekend were different than their weekday. None the stations are in Loop. there were 3 routes that start and end at the same stations, and they were in the recreation and turist attraction area.
The member riders’ top 20 routes for weekday involved 20 stations.
Popular Routes
The maximum number of routes possible is 3,908,529 (1,977 stations x 1,977 stations). There were 252,586 unique routes in the data. The top 20 routes analysis in previous section was account for only 1.94% of total routes. In this section, 2,000 routes, which accounted for around 25% of trips, will be used to visualize the popular routes in the map.
Overall
