Eric Wheeler 2024-03-11
Bellabeat is a high-tech manufacturer of health-focused products for women. Their Bellabeat app provides users with health data related to their activity, sleep, stress, menstrual cycle, and mindfulness habits. The Chief Creative Officer of Bellabeat believes that analyzing smart device fitness data could help unlock new growth opportunities for the company.
Analyze smart device data to gain insight into how consumers are using their smart devices. The insights will then be used to help guide marketing strategy for the company.
What are some trends in smart device usage?
How could these trends apply to Bellabeat customers?
How could these trends help influence Bellabeat marketing strategy?
FitBit Fitness Tracker Data (CC0: Public Domain, dataset made available through Mobius): This Kaggle data set contains personal fitness tracker from thirty Fitbit users. Thirty eligible Fitbit users consented to the submission of personal tracker data, including minute-level output for physical activity, heart rate, and sleep monitoring. It includes information about daily activity, steps, and heart rate that can be used to explore users’ habits.
I have identified a number of limitations to be aware of for this dataset:
Small sample size - there are only 30 participants in this dataset
Short time frame - the data only covers a period of two months which could lead to possible seasonal influences on the level of activity performed by participants.
Inconsistent data - there are gaps in the data such as not all participants using the fitness trackers while they sleep, and weight data is a mixture of self-reported and auto-logged measurements.
The Bellabeat dataset contains 2 folders that have a total of 29 CSV files.
One folder consists of 11 CSV files for the 3.12.16 - 4.11.16 date range.
The other folder consists of 18 CSV files for the 4.12.16 - 5.12.16 date range.
The file names in the two folders are identical, which could lead to confusion when trying to work with them. After selecting the files that would be useful for this analysis, I updated the file names to include the date range for the data.
I chose the files that contained:
Daily Activity
Hourly Steps
Minute Sleep
Sleep Day
Each of these areas had corresponding files in each folder, with the exception of sleep data. I will use data manipulation to convert the Minute Sleep data to Sleep Day data.
Load the necessary packages and import the CSV files.
library(tidyverse) # this collection contains packages used in the analysis i.e. dplyr, ggplot2, tidyr, lubridate## ── Attaching core tidyverse packages ──────────────────────── tidyverse 2.0.0 ──
## ✔ dplyr 1.1.4 ✔ readr 2.1.5
## ✔ forcats 1.0.0 ✔ stringr 1.5.1
## ✔ ggplot2 3.5.0 ✔ tibble 3.2.1
## ✔ lubridate 1.9.3 ✔ tidyr 1.3.1
## ✔ purrr 1.0.2
## ── Conflicts ────────────────────────────────────────── tidyverse_conflicts() ──
## ✖ dplyr::filter() masks stats::filter()
## ✖ dplyr::lag() masks stats::lag()
## ℹ Use the conflicted package (<http://conflicted.r-lib.org/>) to force all conflicts to become errorsdaily_activity_1 <- read.csv("C:/Users/Eric/OneDrive/Desktop/Bellabeat Folder/3.12.16-4.11.16_dailyActivity_merged.csv")
hourly_steps_1 <- read.csv("C:/Users/Eric/OneDrive/Desktop/Bellabeat Folder/3.12.16-4.11.16_hourlySteps_merged.csv")
minute_sleep_1 <- read.csv("C:/Users/Eric/OneDrive/Desktop/Bellabeat Folder/3.12.16-4.11.16_minuteSleep_merged.csv")
daily_activity_2 <- read.csv("C:/Users/Eric/OneDrive/Desktop/Bellabeat Folder/4.12.16-5.12.16_dailyActivity_merged.csv")
hourly_steps_2 <- read.csv("C:/Users/Eric/OneDrive/Desktop/Bellabeat Folder/4.12.16-5.12.16_hourlySteps_merged.csv")
day_sleep_2 <- read.csv("C:/Users/Eric/OneDrive/Desktop/Bellabeat Folder/4.12.16-5.12.16_sleepDay_merged.csv")Let’s validate the number of participants in each dataframe.
## [1] 35## [1] 33## [1] 23## [1] 24## [1] 34## [1] 33There is fluctuating participation for each dataset as well as more participants than stated in the brief. Also, it appears that there were much fewer participants using their smart devices to track their sleep.
I’ve already identified that one dataset on sleep contains minute measurements while another contains daily measurements for sleep minutes. First I’ll look at both sleep files to see how I can approach aligning the data. Then I’ll make the necessary adjustments to both data frames to prepare them to be combined later on.
## Rows: 198,559
## Columns: 4
## $ Id <dbl> 1503960366, 1503960366, 1503960366, 1503960366, 1503960366, 1503…
## $ date <chr> "3/13/2016 02:39:30", "3/13/2016 02:40:30", "3/13/2016 02:41:30"…
## $ value <int> 1, 1, 1, 1, 1, 1, 2, 2, 1, 1, 2, 2, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1…
## $ logId <dbl> 11114919637, 11114919637, 11114919637, 11114919637, 11114919637,…## Rows: 413
## Columns: 5
## $ Id <dbl> 1503960366, 1503960366, 1503960366, 1503960366, 150…
## $ SleepDay <chr> "4/12/2016", "4/13/2016", "4/15/2016", "4/16/2016",…
## $ TotalSleepRecords <int> 1, 2, 1, 2, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, …
## $ TotalMinutesAsleep <int> 327, 384, 412, 340, 700, 304, 360, 325, 361, 430, 2…
## $ TotalTimeInBed <int> 346, 407, 442, 367, 712, 320, 377, 364, 384, 449, 3…# Make adjustments to minute_sleep_1
minute_sleep_1 <- minute_sleep_1[-4] %>% # Remove the logId column
mutate(date = as.Date(date, format = "%m/%d/%Y %H:%M:%S")) %>% # Change the date data type to timestamp
rename(Date = date) # Rename the date column to align with day_sleep_2
# Create a new aggregated day_sleep_1 dataframe
day_sleep_1 <- minute_sleep_1 %>%
group_by(Id, Date) %>% # group the data by the Id and Date columns
summarise(TotalMinutesAsleep = sum(value)) # Get the sum of the minute sleep values## `summarise()` has grouped output by 'Id'. You can override using the `.groups`
## argument.# Adjust day_sleep_2 to match newly created day_sleep_1
day_sleep_2 <- day_sleep_2[-c(3,5)] %>% # Drop unneccessary columns
mutate(SleepDay = as.Date(SleepDay, format = "%m/%d/%Y")) %>% # Change the SleepDay data type to timestamp
rename(Date = SleepDay) # rename the SleepDay column## Id ActivityDate TotalSteps
## 0 0 0
## TotalDistance TrackerDistance LoggedActivitiesDistance
## 0 0 0
## VeryActiveDistance ModeratelyActiveDistance LightActiveDistance
## 0 0 0
## SedentaryActiveDistance VeryActiveMinutes FairlyActiveMinutes
## 0 0 0
## LightlyActiveMinutes SedentaryMinutes Calories
## 0 0 0## Id ActivityDate TotalSteps
## 0 0 0
## TotalDistance TrackerDistance LoggedActivitiesDistance
## 0 0 0
## VeryActiveDistance ModeratelyActiveDistance LightActiveDistance
## 0 0 0
## SedentaryActiveDistance VeryActiveMinutes FairlyActiveMinutes
## 0 0 0
## LightlyActiveMinutes SedentaryMinutes Calories
## 0 0 0## Id Date TotalMinutesAsleep
## 0 0 0## Id Date TotalMinutesAsleep
## 0 0 0## Id ActivityHour StepTotal
## 0 0 0## Id ActivityHour StepTotal
## 0 0 0There are no null values in any of the data frames.
We can union the tables to combine the 3/12/16-4/11/16 data with the 4/12/16-5/12/16 data. Once this is done we continue cleaning by adjusting data types and removing duplicates
# Combine the sleep data
day_sleep_df <- rbind(day_sleep_1, day_sleep_2)
# combine the daily activity data
daily_activity_df <- rbind(daily_activity_1, daily_activity_2)
# drop distance related columns in the daily acivity data
daily_activity_df <- daily_activity_df[-c(4:10)]
# make adjustments to the daily activity data
daily_activity_df <- daily_activity_df %>%
mutate(ActivityDate = as.Date(ActivityDate, format = "%m/%d/%Y")) %>% # Change ActivityDate data type to Date
rename(Date = ActivityDate) # rename the ActivityDate column
# combine the hourly steps data
hourly_steps_df <- rbind(hourly_steps_1,hourly_steps_2)
# make adjustments to the steps data
hourly_steps_df$ActivityHour <- strptime(hourly_steps_df$ActivityHour, format="%m/%d/%Y %H:%M:%S") # change the ActivityHour data type to timestampNow remove any duplicate rows that might exist in the data.
daily_activity_df <- daily_activity_df %>% distinct()
day_sleep_df <- day_sleep_df %>% distinct()
hourly_steps_df <- hourly_steps_df %>% distinct()In order to uncover some insights from the data it will be helpful to create some features from the existing dimensions. In terms of date/time columns, we can extract the hour of the day, day of the week, and week of the year from the data.
# get the WeekOfYear and DayOfWeek for the daily activity data
daily_activity_df$WeekOfYear <- as.integer(week(daily_activity_df$Date)) # create a WeekOfYear column
daily_activity_df$DayOfWeek <- wday(daily_activity_df$Date, label=TRUE, abbr=FALSE) # create a DayOfWeek column
# get the DayOfWeek for the sleep data
day_sleep_df$DayOfWeek <- wday(day_sleep_df$Date, label=TRUE, abbr=FALSE)
# get the HourOfDay for hourly steps data
hourly_steps_df$HourOfDay <- hour(hourly_steps_df$ActivityHour)It would also be helpful to create some benchmarks that are fairly universal regardless of age, sex, or race. We can create bins for steps per day and sleep per day.
We can create sleep categories that group users based on this information.
# Define conditions and create SleepCategory column
day_sleep_df <- day_sleep_df %>%
mutate(SleepCategory = case_when(
TotalMinutesAsleep >= 420 & TotalMinutesAsleep <= 540 ~ "7 to 9 hours",
TotalMinutesAsleep < 420 ~ "less than 7 hours",
TotalMinutesAsleep > 540 ~ "more than 9 hours",
TRUE ~ NA_character_ # for any other cases
))
# adjust the categories into a specified order
sleep_categories <- c("less than 7 hours", "7 to 9 hours", "more than 9 hours")
# set the categorical variables
day_sleep_df$SleepCategory <- factor(day_sleep_df$SleepCategory, levels = sleep_categories)Sedentary is equal to <5000 steps/day
Low Active is equal to 5000-7499 steps/day
Somewhat Active is equal to 7500-9999 steps/day
Active is equal to >=10000 steps/day
Very Active is equal to >12500 steps/day
We can create these categories as well.
# Define conditions and create ActivityLevel column
daily_activity_df <- daily_activity_df %>%
mutate(ActivityLevel = case_when(
is.na(TotalSteps) ~ "Unknown",
TotalSteps < 5000 ~ "Sedentary",
TotalSteps < 7500 ~ "Low Active",
TotalSteps < 10000 ~ "Somewhat Active",
TotalSteps < 12500 ~ "Active",
TRUE ~ "Very Active"
))
# adjust the categories into a specified order
activity_levels <- c("Sedentary", "Low Active", "Somewhat Active", "Active", "Very Active")
# set the categorical variables
daily_activity_df$ActivityLevel <- factor(daily_activity_df$ActivityLevel, levels = activity_levels)Get the summary statistics for our dataframes to understand the data better.
## Id Date TotalSteps VeryActiveMinutes
## Min. :1.504e+09 Min. :2016-03-12 Min. : 0 Min. : 0.00
## 1st Qu.:2.320e+09 1st Qu.:2016-04-09 1st Qu.: 3146 1st Qu.: 0.00
## Median :4.445e+09 Median :2016-04-19 Median : 6999 Median : 2.00
## Mean :4.781e+09 Mean :2016-04-19 Mean : 7281 Mean : 19.68
## 3rd Qu.:6.962e+09 3rd Qu.:2016-04-30 3rd Qu.:10544 3rd Qu.: 30.00
## Max. :8.878e+09 Max. :2016-05-12 Max. :36019 Max. :210.00
##
## FairlyActiveMinutes LightlyActiveMinutes SedentaryMinutes Calories
## Min. : 0.0 Min. : 0.0 Min. : 0.0 Min. : 0
## 1st Qu.: 0.0 1st Qu.:111.0 1st Qu.: 729.0 1st Qu.:1799
## Median : 6.0 Median :195.0 Median :1057.0 Median :2114
## Mean : 13.4 Mean :185.4 Mean : 992.5 Mean :2266
## 3rd Qu.: 18.0 3rd Qu.:262.0 3rd Qu.:1244.0 3rd Qu.:2770
## Max. :660.0 Max. :720.0 Max. :1440.0 Max. :4900
##
## WeekOfYear DayOfWeek ActivityLevel
## Min. :11.00 Sunday :193 Sedentary :500
## 1st Qu.:15.00 Monday :188 Low Active :254
## Median :16.00 Tuesday :225 Somewhat Active:213
## Mean :16.15 Wednesday:198 Active :231
## 3rd Qu.:18.00 Thursday :195 Very Active :199
## Max. :19.00 Friday :199
## Saturday :199## Id Date TotalMinutesAsleep DayOfWeek
## Min. :1.504e+09 Min. :2016-03-11 Min. : 6.0 Sunday :129
## 1st Qu.:3.977e+09 1st Qu.:2016-03-27 1st Qu.: 379.0 Monday :119
## Median :4.703e+09 Median :2016-04-10 Median : 454.0 Tuesday :131
## Mean :4.910e+09 Mean :2016-04-10 Mean : 441.8 Wednesday:125
## 3rd Qu.:6.776e+09 3rd Qu.:2016-04-26 3rd Qu.: 526.0 Thursday :123
## Max. :8.792e+09 Max. :2016-05-12 Max. :1201.0 Friday :121
## Saturday :129
## SleepCategory
## less than 7 hours:323
## 7 to 9 hours :369
## more than 9 hours:185
##
##
##
## ## Id ActivityHour StepTotal
## Min. :1.504e+09 Min. :2016-03-12 00:00:00.00 Min. : 0.0
## 1st Qu.:2.320e+09 1st Qu.:2016-03-26 13:00:00.00 1st Qu.: 0.0
## Median :4.559e+09 Median :2016-04-10 11:00:00.00 Median : 21.0
## Mean :4.870e+09 Mean :2016-04-10 17:33:39.75 Mean : 302.9
## 3rd Qu.:6.962e+09 3rd Qu.:2016-04-25 16:00:00.00 3rd Qu.: 323.0
## Max. :8.878e+09 Max. :2016-05-12 15:00:00.00 Max. :10565.0
## NA's :33
## HourOfDay
## Min. : 0.00
## 1st Qu.: 5.00
## Median :11.00
## Mean :11.45
## 3rd Qu.:17.00
## Max. :23.00
## Summary Statistics Insights:
The mean and median daily steps for these participants would be considered “low active”
The average daily sedentary time for participants is over 16 hours a day. That’s a lot of sedentary time!
The average sleep per night is between 7 and 8 hours.
The median steps taken per hour is a mere 21 steps!
I’ll start by looking at the hourly breakdown of steps for these participants
ggplot(data=hourly_steps_df, mapping = aes(x=HourOfDay, y=StepTotal, fill=HourOfDay)) +
geom_bar(stat="sum") +
theme_minimal() +
theme(legend.position = "none") +
labs(title = "Total Steps by Hour of the Day")
Hourly Steps Insights:
It looks like the most steps among these participants were taken from 6-7 PM. Participants were the most active possibly just before or just after dinner time.
During a typical 9-5 workday, the most steps were taken at 12 and 2. Generally speaking, it appears that these participants tended to do the most walking right around mealtimes.
# calculate the percentage of each ActivityLevel
activity_percentage <- daily_activity_df %>%
count(ActivityLevel) %>%
mutate(percentage = n / sum(n) * 100)
# Create the pie chart
ggplot(activity_percentage, aes(x = "", y = n, fill = ActivityLevel)) +
geom_bar(stat = "identity") +
coord_polar("y", start = 0) +
geom_text(aes(label = paste0(round(percentage), "%")), position = position_stack(vjust = 0.5)) +
labs(title = "Activity Level Distribution") +
theme_void()
Activity Level Insights:
# calculate the percentage of each ActivityLevel by DayOfWeek
dow_activity <- daily_activity_df %>%
group_by(DayOfWeek, ActivityLevel) %>%
summarise(count = n()) %>%
group_by(DayOfWeek) %>%
mutate(percentage = count / sum(count) * 100)## `summarise()` has grouped output by 'DayOfWeek'. You can override using the
## `.groups` argument.# create a bar chart
ggplot(dow_activity, aes(x=DayOfWeek, y=count, fill=ActivityLevel)) +
geom_bar(stat = "identity") +
geom_text(aes(label = paste0(round(percentage), "%")),
position = position_stack(vjust = 0.5), size = 3) +
labs(title="Activity Level by Day of Week", x = "Day of Week", y = "Count", fill = "Activity Level") +
theme(axis.text.x = element_text(angle = 45, hjust = 1)) +
theme_minimal()
Day of Week Activity Level Insights:
# calculate percentage breakdown of Sleep Quality
sleep_percentage <- day_sleep_df %>% group_by(SleepCategory) %>% summarize(count = n()) %>% mutate(percentage = count / sum(count) * 100)
# Create pie chart
ggplot(sleep_percentage, aes(x = "", y = percentage, fill = SleepCategory)) +
geom_bar(stat = "identity") +
coord_polar("y", start = 0) +
theme_void() + theme(legend.position = "bottom") +
geom_text(aes(label = paste0(round(percentage), "%")), position = position_stack(vjust = 0.5)) +
labs(title = "Sleep Quality Breakdown", fill = "Sleep Quality") +
theme_void()
Sleep insights
37% of days users were getting less than 7 hours of sleep
Less than 50% of users were getting the appropriate 7-9 hours of sleep
# calculate counts of SleepCategory for each DayOfWeek
sleep_counts <- day_sleep_df %>%
group_by(DayOfWeek, SleepCategory) %>%
summarise(count = n()) %>%
group_by(DayOfWeek) %>%
mutate(percentage = count / sum(count) * 100)## `summarise()` has grouped output by 'DayOfWeek'. You can override using the
## `.groups` argument.# create the bar chart
ggplot(sleep_counts, aes(x = DayOfWeek, y = count, fill = SleepCategory)) +
geom_bar(stat = "identity") +
geom_text(aes(label = paste0(round(percentage), "%")),
position = position_stack(vjust = 0.5), size = 3) +
labs(title="Sleep Quality by Day of Week", x = "Day of Week", y = "Count", fill = "Sleep Category") +
theme(axis.text.x = element_text(angle = 45, hjust = 1)) +
theme_minimal()
Day of the Week Sleep Insights:
Sundays and Saturdays were the days most often for users to get more than 9 hours of sleep. Considering that these were also the days with the highest likelihood of being considered sedentary, we should look to see if there’s a correlation between sleep and sedentary minutes and activity levels.
During the week it’s likely that participants are getting the appropriate 7-9 hours of sleep less than 50% of the time. In fact, on days like Tuesday and Friday, around 50% of nights had less than 7 hours of sleep.
Remove days with double log entries in the daily_activity_df before joining
# find rows that contain the same Id and Date i.e. double entries
duplicate_rows <- daily_activity_df %>%
filter(duplicated(daily_activity_df[, c("Id", "Date")]) | duplicated(daily_activity_df[, c("Id", "Date")], fromLast = TRUE))
# Remove double entries from daily_activity_df
daily_activity_df <- anti_join(daily_activity_df, duplicate_rows, by = c("Id", "Date"))Join the Daily Activity and Day Sleep tables to see rows that have both sleep and activity data
# create a scatter plot for TotalMinutesAsleep and SedentaryMinutes with facets around SleepCategory
ggplot(data=daily_df, mapping = aes(x=TotalMinutesAsleep, y=SedentaryMinutes)) +
geom_point() +
geom_smooth() +
theme_minimal() +
facet_wrap(~SleepCategory) +
labs(title = "Correlation Between Minutes Asleep and Minutes Spent Sedentary")## `geom_smooth()` using method = 'loess' and formula = 'y ~ x'
Sleep and Sedentary Minutes Insights:
# calculate counts of instances for each combination of SleepCategory and ActivityLevel
count_data <- daily_df %>%
group_by(SleepCategory, ActivityLevel) %>%
summarise(count = n()) %>%
mutate(percentage = count / sum(count) * 100)## `summarise()` has grouped output by 'SleepCategory'. You can override using the
## `.groups` argument.# Create the bar chart
ggplot(count_data, aes(x = SleepCategory, y = count, fill = ActivityLevel)) +
geom_bar(stat = "identity", position = "stack") +
geom_text(aes(label = paste0(round(percentage), "%")), position = position_stack(vjust = 0.5), size = 3) +
labs(title="Breakdown of Activity Level by Sleep Category", x = "Sleep Category", y = "Count", fill = "Activity Level") +
theme(axis.text.x = element_text(angle = 45, hjust = 1)) +
theme_minimal()
Activity Level and Sleep Category Insights:
Surprisingly, days where there participants had less than 7 hours of sleep actually produced the lowest instances of sedentary days and the highest level of very active days.
Days where participants slept more than 9 hours were also the most likely to be labeled as sedentary
Only 31% of days hit the recommended daily step count of 10K steps, with the average daily step count putting these participants in the “low active” category. The most common Activity Level in this study was “sedentary,” with the average daily sedentary time coming in at over 16 hours a day. By using the Activity Level Benchmarks provided by the NIH, the smart devices can update users on what category they are currently in, and provide them with a countdown to how many steps will get them to the next Activity Level.
The most steps taken each day fell around meal times. A study published in the Springer Journal of Sports Medicine claims that walking after eating regulates blood sugar, improves blood circulation, and lowers blood pressure. By adding a feature on the smart device that monitors sedentary time between 12-2PM and 6-7PM, users can be reminded to get active after a meal to receive these benefits.
Less than 50% of days consisted of the appropriate 7-9 hours of sleep and 37% of days had participants getting less than 7 hours of sleep. Getting more sleep was correlated with participants spending fewer minutes sedentary that day. If users are wearing their smart devices to track sleep, the device can give warnings if the user is approaching the “less than 7 hours” sleep category.
Participants were also sleeping the most and walking the least on Sundays, while Saturday also had a higher likelihood to be considered “sedentary” compared with weekdays. The smart device could have special settings for weekends to remind users to get active if sedentary time hits a certain threshold.