Bellabeat Analysis (2016)

Objective

Analyze FitBit fitness tracker data from 30 users to understand activity patterns, sleep behavior, and feature utilization to inform Bellabeat's marketing strategy and product development.

Tools

R: tidyverse, dplyr, ggplot2, lubridate, forcats, vroom, janitor, here, readr, scales, broom
Visualization: ggplot2, heatmaps, scatter plots, boxplots, bar charts, line charts

Project Structure

This outlines the main directories and files used in the Bellabeat analysis pipeline.

Bellabeat-Analysis/
├── data/
│   ├── raw/                  # Raw CSV files (18 files, Input Data)
│   ├── cleaned/              # Cleaned individual datasets (18 files)
│   ├── processed/            # Analysis output CSVs (Input for visualization.R)
│   └── output/               # Exploration outputs (optional)
├── plots/                    # Generated PNG visualizations (20 plots)
│   ├── 01_users_per_dataset.png
│   ├── 02_engagement_distribution.png
│   ├── 03_activity_level_distribution.png
│   ├── 04_feature_utilization.png
│   ├── 05_activity_heatmap.png
│   ├── 06_peak_activity_hours.png
│   ├── 07_steps_by_activity_segment.png
│   ├── 08_weekday_vs_weekend.png
│   ├── 09_correlation_matrix.png
│   ├── 10_steps_vs_calories.png
│   ├── 11_active_minutes_vs_calories.png
│   ├── 12_key_correlations.png
│   ├── 13_weekly_trends.png
│   ├── 14_user_retention.png
│   ├── 15_day_of_week_patterns.png
│   ├── 16_sleep_duration_histogram.png
│   ├── 17_sleep_efficiency_by_segment.png
│   ├── 18_time_in_bed_vs_asleep.png
│   ├── 19_steps_distribution_with_goal.png
│   └── 20_business_insights_summary.png
├── scripts/                  # Core R scripts
│   ├── 01_data_cleaning.R    # Cleans and combines all 18 raw CSVs
│   ├── 02_exploration.R      # Data coverage, quality checks, basic distributions
│   ├── 03_analysis.R         # User segmentation, correlations, temporal patterns
│   └── 04_visualization.R    # Generates all 20 plots from processed data
├── analysis_guide.Rmd        # Comprehensive analysis guide and methodology
├── README.md                 # This file
└── .gitignore                # Specifies files and folders to ignore

Process

1. Data Cleaning (`scripts/01_data_cleaning.R`)

Load all 18 CSV files from data/raw/
Clean column names using janitor::clean_names()
Parse dates with custom datetime parser
Convert IDs to character type
Remove invalid data:
- Heart rates >220 bpm or ≤0
- Total minutes >1440 per day
- Duplicate records
Round distance measurements to 2 decimal places
Calculate derived metrics (total_active_minutes, active_ratio, non_wear flags)
Save 18 cleaned datasets to data/cleaned/

2. Exploration (`scripts/02_exploration.R`)

A. Data Coverage & Quality

Count users per dataset
Calculate tracking days per user (range: 4-31 days)
Classify engagement levels: Everyday (25-31), Heavy (21-24), Moderate (11-20), Light (1-10 days)
Identify which users have sleep, heart rate, and weight data

B. Basic Distributions

Summary statistics: steps, calories, active minutes, sleep duration
Outlier detection: zero-step days, extreme calories, abnormal sleep

C. Feature Utilization

Dataset availability per user (activity, sleep, heart rate, weight)
Tracking frequency and consistency
Feature adoption rates

Outputs: users_per_dataset.csv, user_engagement.csv, feature_summary.csv, feature_utilization_matrix.csv

3. Analysis (`scripts/03_analysis.R`)

A. User Segmentation

Activity levels: Sedentary (<5k), Low Active (5k-7.5k), Somewhat Active (7.5k-10k), Active (10k-12.5k), Highly Active (≥12.5k steps)
Sleep patterns: Under-sleeper (<6h), Normal (6-8h), Over-sleeper (>8h)

B. Correlation Analysis

Steps → Calories (strong positive expected)
Very Active Minutes → Calories
Sedentary Minutes → Calories (weak/negative expected)
Sleep Duration → Next Day Steps (lag analysis)
Sleep Efficiency correlations

C. Temporal Patterns

Hourly activity summary (peak hours: likely 5 AM - 9 PM)
Day of week patterns (weekday vs weekend)
Weekly trends over 2-month period
User retention analysis

D. Sleep Analysis

Sleep efficiency (time asleep / time in bed)
Sleep consistency (standard deviation)
Relationship between activity and sleep quality

E. Business Questions

Peak activity times for notification timing
% users meeting 10k step goal
% users getting adequate sleep (7-9 hours)
Feature adoption by activity segment
User retention and drop-off rates

Outputs: user_segments.csv, correlation_matrix.csv, key_correlations.csv, hourly_activity_summary.csv, daily_averages_by_dow.csv, weekly_trends.csv, sleep_analysis.csv, business_insights.csv

4. Visualization (`scripts/04_visualization.R`)

A. User Overview Dashboard

Bar chart: users per dataset
Pie chart: engagement distribution
Bar chart: activity level distribution
Stacked bar: feature utilization

B. Activity Patterns

Heatmap: hour × day of week activity
Line chart: peak activity hours
Boxplot: steps by activity segment
Bar chart: weekday vs weekend comparison

C. Correlations

Heatmap: correlation matrix
Scatter plots with regression: steps vs calories, active minutes vs calories
Bar chart: key correlations summary

D. Temporal Trends

Line chart: weekly activity trends
Bar chart: user retention funnel
Line chart: day of week patterns

E. Sleep Visualizations

Histogram: sleep duration with 7-9h guidelines
Boxplot: sleep efficiency by activity level
Scatter: time in bed vs time asleep

F. Business Insights

Histogram: steps distribution with 10k goal line
Text visual: key business insights summary

Key Insights

User Behavior

30 users tracked between March 12 - May 12, 2016 (2 months)
Tracking consistency varies: 4-31 days per user
Feature adoption:
- Activity: 100% (all users)
- Sleep: ~80% of users
- Heart rate: ~47% of users
- Weight: ~27% of users (very sparse)

Activity Patterns

Users more active during weekdays (commuting patterns)
Peak activity hours: likely 12 PM - 7 PM
Strong correlation between steps and calories (r > 0.7 expected)
Very active minutes drive calorie burn more than total steps

Sleep Insights

Average sleep: ~7 hours per night
~30% of time spent sleeping, ~50% sedentary
Sleep efficiency varies by activity level
Relationship between daily activity and sleep quality

Business Recommendations

Notification timing: Send activity reminders during peak hours
10k step goal: Track what % of users consistently meet this goal
Sleep tracking: High adoption (80%) suggests users value this feature
Weight tracking: Low adoption (27%) - consider simplifying or deprioritizing
Retention: Monitor user drop-off rates week-over-week

Dataset Limitations

Small sample size (30 users) limits generalizability
Dataset from 2016 is outdated; trends may have changed
No demographic information (age, gender, location)
Only 1-2 months of data; cannot capture long-term trends
Not all users tracked all features consistently
Tracking days vary widely (4-31 days)

Deliverables

Cleaned Data: 18 cleaned CSV files in data/cleaned/
Analysis Outputs: 13+ processed CSVs in data/processed/
Visualizations: 20 publication-ready plots in plots/
Scripts:
Documentation: Analysis Guide

Example plots:

Generating the Analysis Locally

The raw data is available on Kaggle. To reproduce the analysis:

Download raw data from Kaggle into data/raw/
- Ensure all 18 CSV files are present
Run data cleaning script
```
source("scripts/01_data_cleaning.R")
```
- Outputs: 18 cleaned CSVs in data/cleaned/
Run exploration script
```
source("scripts/02_exploration.R")
```
- Outputs: Coverage and quality reports in data/processed/
Run analysis script
```
source("scripts/03_analysis.R")
```
- Outputs: Analysis CSVs in data/processed/
Generate visualizations
```
source("scripts/04_visualization.R")
```
- Outputs: 20 plots in plots/ folder

Required R Packages

install.packages(c(
  "vroom", "dplyr", "readr", "here", "lubridate",
  "tidyr", "janitor", "ggplot2", "forcats", "scales", "broom"
))