Bike-Share Analysis 2019-2020

Goals:

1. Calculate the average trip duration for members (Subscribers) and casual riders (Customers)
2. Compare difference in usage between years
3. Calculate the number of trips & average trip length for different user types on different days of the week

N.B. * In the data, members are referred to as a “Subscriber” and casual riders are referred to as a “Customer”.

1. Calculate the average trip_duration for members (Subscribers) and casual riders (Customers) & compare across years

Summarize mean_trip_duration and group by user_type

bike_usage_usertype_avg <- bike_trips_2019_2020 %>% 
  group_by(user_type) %>% 
  drop_na() %>% 
  summarize(mean_trip_duration = mean(trip_duration), .groups = "drop")

bike_usage_usertype_avg %>%
  rename(
    User_Type = user_type,
    Mean_Trip_Duration = mean_trip_duration
  ) %>%
  kable(caption = "Table 1: Mean Trip Duration by User Type") %>%
  kable_styling(full_width = FALSE) %>%
  scroll_box(width = "100%")

Table 1: Mean Trip Duration by User Type

User_Type	Mean_Trip_Duration
Customer	1822.938
Subscriber	659.095

ggplot(bike_usage_usertype_avg, aes(x = user_type, y = mean_trip_duration / 60, fill = user_type)) +
  geom_col(width = 0.6) +
  labs(title = "Graph 1: Average Trip Duration by User Type",
       x = "User Type", y = "Mean Trip Duration (minutes)") +
  scale_fill_manual(values = c("Customer" = "#E69F00", "Subscriber" = "#56B4E9")) +
  theme_minimal()

Key Takeaway:

• Customers take significantly longer trips on average - almost 3x as long as subscribers.

Possible reasons:

• Casual riders use bikes for leisure rides, which tend to be longer
• Subscribers use bikes for utilitarian trips (commuting, errands), which tend to be shorter, but more often worked into everyday life (routine, planned)

2. Compare difference in usage between years

Summarize mean_trip_duration and group by user_type and year to compare

bike_usage_usertype_year_avg <- bike_trips_2019_2020_v2 %>% 
  group_by(user_type, year) %>% 
  drop_na() %>% 
  summarize(mean_trip_duration = mean(trip_duration), .groups = "drop")

bike_usage_usertype_year_avg %>%
  rename(
    User_Type = user_type,
    Year = year,
    Mean_Trip_Duration = mean_trip_duration
  ) %>%
  kable(caption = "Table 2: Mean Trip Duration by User Type and Year") %>%
  kable_styling(full_width = FALSE) %>%
  scroll_box(width = "100%")

Table 2: Mean Trip Duration by User Type and Year

User_Type	Year	Mean_Trip_Duration
Customer	2019	1911.8329
Customer	2020	1780.0174
Subscriber	2019	648.7644
Subscriber	2020	668.4237

bike_usage_usertype_year_avg$year <- as.factor(bike_usage_usertype_year_avg$year)

ggplot(bike_usage_usertype_year_avg, aes(x = year, y = mean_trip_duration / 60, fill = user_type)) +
  geom_col(position = "dodge") +
  labs(title = "Graph 2: Average Trip Duration by Year and User Type",
       x = "Year", y = "Mean Trip Duration (minutes)") +
  scale_fill_manual(values = c("Customer" = "#E69F00", "Subscriber" = "#56B4E9")) +
  theme_minimal()

Key Takeaways:

• Customers’ greater usage of the service is consistent across both years
• Between 2019 and 2020 there was:
  • A minor decrease in mean trip duration for customers (by ~132s)
  • A negligible increase in mean trip duration for subscribers (by ~19s)

3. Calculate the number of trips & average trip length for different user types on different days of the week

Count number of rides for each day of the week, calculate the average ride_length for users by each day of the week

bike_usage_usertype_dow_count_avg <- bike_trips_2019_2020_v2 %>% 
  group_by(user_type, day_of_week) %>% 
  drop_na() %>% 
  summarize(trip_count = n(), mean_trip_duration = mean(trip_duration), .groups = "drop")

Use pivot_wider() and kable() to make table more readable, use mutate() to add columns for total usage minutes for usertype to compare

bike_usage_usertype_dow_count_avg_wide3 <- bike_usage_usertype_dow_count_avg %>%
  pivot_wider(
    names_from = user_type,
    values_from = c(trip_count, mean_trip_duration),
    names_sep = "_"
  ) %>%
  mutate(
    across(starts_with("trip_count_"), round),
    across(starts_with("mean_trip_duration_"), round)
  ) %>%
  mutate(
    total_minutes_Customer = trip_count_Customer * mean_trip_duration_Customer,
    total_minutes_Subscriber = trip_count_Subscriber * mean_trip_duration_Subscriber
  )

bike_usage_usertype_dow_count_avg_wide3 %>%
  rename(
    Day = day_of_week,
    Customer_Trips = trip_count_Customer,
    Subscriber_Trips = trip_count_Subscriber,
    Customer_Avg_Sec = mean_trip_duration_Customer,
    Subscriber_Avg_Sec = mean_trip_duration_Subscriber,
    Customer_Total_Sec = total_minutes_Customer,
    Subscriber_Total_Sec = total_minutes_Subscriber
  ) %>%
  kable(caption = "Table 3: Bike Usage Metrics by User Type and Day of Week") %>%
  kable_styling(full_width = FALSE) %>%
  scroll_box(width = "100%")

Table 3: Bike Usage Metrics by User Type and Day of Week

Day	Customer_Trips	Subscriber_Trips	Customer_Avg_Sec	Subscriber_Avg_Sec	Customer_Total_Sec	Subscriber_Total_Sec
Monday	6632	110350	1353	643	8973096	70955050
Tuesday	7881	127861	1572	650	12388932	83109650
Wednesday	8254	121816	1867	650	15410218	79180400
Thursday	7665	125137	1549	643	11873085	80463091
Friday	8378	115056	1618	637	13555604	73290672
Saturday	13297	59333	1984	709	26381248	42067097
Sunday	18434	60146	2170	754	40001780	45350084

ggplot(bike_usage_usertype_dow_count_avg, aes(x = day_of_week, y = trip_count, fill = user_type)) +
  geom_col(position = "dodge") +
  labs(
    title = "Graph 3: Trip Count by Day of Week and User Type",
    x = "Day of Week", y = "Trip Count", fill = "User Type"
  ) +
  scale_y_continuous(labels = scales::label_comma()) +
  scale_fill_manual(values = c("Customer" = "#E69F00", "Subscriber" = "#56B4E9")) +
  theme_minimal()

bike_usage_usertype_dow_count_avg_wide3 %>%
  rename(
    Day = day_of_week
  ) %>%
  pivot_longer(cols = c(total_minutes_Customer, total_minutes_Subscriber),
               names_to = "User_Type", values_to = "Total_Minutes") %>%
  mutate(User_Type = ifelse(User_Type == "total_minutes_Customer", "Customer", "Subscriber")) %>%
  ggplot(aes(x = Day, y = Total_Minutes, fill = User_Type)) +
  geom_col(position = "dodge") +
  labs(title = "Graph 4: Total Usage Minutes by Day and User Type",
       x = "Day of Week", y = "Total Minutes") +
  scale_fill_manual(values = c("Customer" = "#E69F00", "Subscriber" = "#56B4E9")) +
  scale_y_continuous(labels = scales::label_comma()) +
  theme_minimal()

bike_usage_usertype_dow_count_avg %>%
  ggplot(aes(x = day_of_week, y = mean_trip_duration / 60, color = user_type, group = user_type)) +
  geom_line(linewidth = 1.2) +
  geom_point(size = 2) +
  labs(title = "Graph 5: Average Trip Duration by Day and User Type",
       x = "Day of Week", y = "Mean Trip Duration (minutes)", color = "User Type") +
  scale_color_manual(values = c("Customer" = "#E69F00", "Subscriber" = "#56B4E9")) +
  theme_minimal()

Key Takeaways:

Average Trip Duration:

• Across all days, customers always ride longer than subscribers.
• Customers ride longest on Sundays (~36 mins), and shortest on Mondays (~22.5 mins)
• Subscribers ride longest on Sundays too (~12.5 mins), and shortest on Fridays (~10.6 mins)
• Usage across both user types peaks on weekends, especially Sundays, when people are more likely to be biking for pleasure

Count:

• Subscribers have a significantly larger quantity of individual/unique rides
  • On weekdays, subscribers have 10x as many individual rides
  • On weekends, subscribers have 3x as many individual rides

Key takeaways (total minutes):

• Although customers have higher average trip durations throughout the week, subscribers have a significantly larger quantity of unique trips which result in a significantly larger amount of total trip minutes
• Members (Subscribers) = high frequency, low duration, higher total minutes = consistent commuting/transport
• Casuals (Customers) = low frequency, high duration, lower total minutes = inconsistent leisure/recreation

Final Conclusions

Key Insights on Consumer Behaviour (Subscribers vs. Customers):

Trip Duration Differences:

• Customers take trips that are, on average, 3x longer than those of Subscribers.
• Suggests Customers use bikes for longer, leisure/recreational purposes.
• Subscribers likely use bikes for shorter, routine, utilitarian trips (commuting, errands).

Trip Frequency:

• Subscribers ride significantly more frequently.
• On weekdays, up to 10x more trips than Customers.
• On weekends, roughly 3x more trips.

Total Usage Time:

• Despite Customers taking longer trips, Subscribers account for significantly more total ride minutes.
• This suggests they are the core base of consistent users — critical for revenue.

Weekday vs. Weekend Behaviour:

• Customers peak on weekends, consistent with recreational use.
• Subscribers ride more evenly throughout the week, especially on weekdays.

Recommendations:

1. Target Weekday-Riding Customers with promotional marketing for subscription model

   • Workweek Plans / Commuter Plans - allow users to save money on regular travel, but encourage more frequent usage

2. Serve repeat-customers prompts to subscribe, persnalized offers based on consumption patterns

   • Customers who take multiple trips within a week, especially on weekdays, are high-conversion prospects
   • Send a notification after a 2nd or 3rd trip in a short period of time, notifying them that they could save on overall costs with a subscription - encourage increased quantity of commuter/errand rides

3. Target recreational weekend customers with promotional material about using bikes during the week, promote commuter usage

   • Promote regular usage and convenience factor

Future Data Collection Suggestions:

• Track ride purpose via optional survey or in-app classification (commute, leisure, errand, etc)
• Gather usage history data of different user types to segment behavior and better personalize future marketing campaigns