OBJECTIVE: To analyze user activity patterns and engagement behavior in Bellabeat fitness tracker data.

KEY INSIGHTS

70% of users (23/33) showed zero engagement with activity logging, while only a small fraction demonstrated consistent usage, indicating a significant engagement gap within the user base.
Consistency in meeting activity thresholds strongly correlates with higher total movement (step count).
Higher consistency is associated with significantly greater time spent in moderate-to-high intensity activity, while light activity remains relatively similar across groups.
Consistently active users exhibit ~3 hours lower sedentary time compared to less active groups, indicating a meaningful behavioral difference.

DATA CLEANING

INSTALLING PAKAGES

library(tidyverse)
library(rmarkdown)

DATA IMPORT

dailyActivity_March_April <- read_csv("mturkfitbit_export_3.12.16-4.11.16/Fitabase Data 3.12.16-4.11.16/dailyActivity_merged.csv")

dailyActivity_April_May <- read_csv("mturkfitbit_export_4.12.16-5.12.16/Fitabase Data 4.12.16-5.12.16/dailyActivity_merged.csv")

COMBINE DATA

Data from two months (March-April and April-May) are combined into one file Uncommon ID’s are removed

common_ids <- intersect(dailyActivity_March_April$Id,
                 dailyActivity_April_May$Id)

dailyActivity_March_April_IDCleaned <- dailyActivity_March_April %>% 
  filter(Id %in% common_ids)

setequal(dailyActivity_March_April_IDCleaned$Id, dailyActivity_April_May$Id) #True

dailyActivity_combined <- bind_rows(dailyActivity_March_April_IDCleaned, dailyActivity_April_May)

CLEAN DATE

Date which is originally formatted as string is formatted to Date

dailyActivity_DateCleaned <- dailyActivity_combined %>% 
  mutate(ActivityDate_cleaned = mdy(dailyActivity_combined$ActivityDate))

max(dailyActivity_DateCleaned$ActivityDate_cleaned) #to check if dates are formatted correctly 
min(dailyActivity_DateCleaned$ActivityDate_cleaned)

DATA ANALYSIS

SEGMENTATION

Users are segmented based on their activity level using Heart Points(HP)

“Heart Points” is a health metric primarily used by Google Fit (developed in collaboration with the American Heart Association and the WHO) to measure physical activity that benefits your heart. Rather than just counting steps, it focuses on intensity. Here is the breakdown:

Moderate Activity: You earn 1 Heart Point for every minute of brisk walking or similar activity where your heart rate increases.
Intense Activity: You earn 2 Heart Points for every minute of higher-intensity activities like running, cycling, or rowing.

The Goal: The standard recommendation is to hit 150 Heart Points per week, which aligns with global health guidelines for reducing the risk of heart disease and improving mental well-being.

STEP1 - people are grouped into 3 groups based on their activity level.

No_of_weeks_with_150_HP	group
>6	Consistently Active
3-5	Ocassionally Active
0-2	Inactive

activity_status <- dailyActivity_DateCleaned %>%
  mutate(week = isoweek(ActivityDate_cleaned)) %>% 
  group_by(Id,week) %>% 
  summarise(total_hp = sum(VeryActiveMinutes)*2 + sum(FairlyActiveMinutes)) %>% 
  arrange(Id,week) %>% 
  mutate(active_week = ifelse(total_hp > 150,1,0)) %>%
  group_by(Id) %>% 
  summarise(active_weeks = sum(active_week), 
            avg_weekly_hp = mean(total_hp), 
            sd_hp = sd(total_hp)
            ) %>%
  mutate(group = case_when(active_weeks >= 6 ~ 'Consistently Active',
                           between(active_weeks,3,5) ~ 'Ocassionally Active',
                           between(active_weeks,0,2) ~ 'Inactive',
                          
                           )
         ) %>% 
   arrange(-active_weeks)
activity_status

STEP 2 - Group level Descriptive Statics is calculated

Group_summary <- dailyActivity_DateCleaned %>% 
  inner_join(activity_status) %>% 
  group_by(group) %>% 
  summarise(across(c(TotalSteps, VeryActiveMinutes, FairlyActiveMinutes, LightlyActiveMinutes, Calories, SedentaryMinutes),
                   list(mean = ~mean(.x),
                        sd = ~sd(.x), 
                        CV = ~sd(.x)/mean(.x)
                        )
                   )
            ) %>% 
  left_join(group_count, by = 'group') %>%  
  arrange(-TotalSteps_mean)
knitr::kable(Group_summary)

group	TotalSteps_mean	TotalSteps_sd	TotalSteps_CV	VeryActiveMinutes_mean	VeryActiveMinutes_sd	VeryActiveMinutes_CV	FairlyActiveMinutes_mean	FairlyActiveMinutes_sd	FairlyActiveMinutes_CV	LightlyActiveMinutes_mean	LightlyActiveMinutes_sd	LightlyActiveMinutes_CV	Calories_mean	Calories_sd	Calories_CV	SedentaryMinutes_mean	SedentaryMinutes_sd	SedentaryMinutes_CV	No_of_people
Consistently Active	11037.886	5037.274	0.4563622	46.122622	38.89035	0.8431948	20.581395	21.11626	1.025988	202.1607	78.30357	0.3873333	2512.562	839.4916	0.3341177	861.277	261.3623	0.3034591	11
Ocassionally Active	6726.277	4137.683	0.6151521	9.580220	16.29173	1.7005586	15.562637	21.75069	1.397622	188.8879	120.79497	0.6395061	2384.066	710.4784	0.2980112	1037.976	306.0314	0.2948348	11
Inactive	4141.190	3701.141	0.8937384	2.811947	10.90912	3.8795612	2.747788	10.43948	3.799232	167.5686	127.96586	0.7636626	1899.836	518.5674	0.2729537	1076.909	319.4748	0.2966590	11

INSIGHTS

Out of 33 individuals only 11 are consistently active.

Group_summary %>% 
  ggplot(aes(x=reorder(group, -No_of_people), y=No_of_people)) + geom_col() + labs(x='group')

TOTAL STEPS

Consistently active individuals are moving around twice then occasionally active and thrice then inactive individuals.

Group_summary %>% 
  ggplot(aes(x=reorder(group, -TotalSteps_mean), y=TotalSteps_mean)) + geom_col() + labs(x='group')

HIGH INTENSITY ACTIVITY

Very active minutes are exponentially high in consistently active individuals.

MODERATE INTENSITY ACTIVITY

Moderate activity levels are exponentially low in Inactive individuals

LOW INTENSITY ACTIVITY

light activity minutes decreased with decreasing activity levels but not much difference

SEDENTARY TIME

sedentary minutes are same of occasionally active and inactive and lower of consistently active

HEART POINTS

Average weekly heart points for consistently active individuals is exponentially high (more then 4 times the recommended minimum activity levels)

Average weekly HP distribution within groups

for consistently active users median of avg weekly HP is around 600 and goes as high as around 1200

activity_status %>% 
  filter(group == 'Consistently Active') %>% 
  ggplot(aes(y=avg_weekly_hp)) + geom_boxplot()

For occasionally active users this median drops to 200 which is more then 150 but they are not consistently reaching 150+ across weeks, some weeks they are more active and some weeks less
Inactive users are not fulfilling the minimum activity guidelines

ENGAGEMENT ANALYSIS

23/33 people didn’t even log distance once

engagement <- dailyActivity_DateCleaned %>% 
  group_by(Id) %>% 
  summarise(No_of_times_logged = sum(ifelse(LoggedActivitiesDistance>0,1,0))) %>% 
  group_by(No_of_times_logged) %>% 
  summarise(No_of_people = n())
knitr::kable(engagement)

No_of_times_logged	No_of_people
0	28
1	2
7	1
19	1
23	1