Goal — Uncover usage patterns (steps, calories, sleep) and translate them into actions that grow engagement and acquisition. Stack — R (tidyverse, dplyr, lubridate, janitor, ggplot2), Tableau Public. Data — 30-day Fitbit dataset (≈30 users, anonymized public data).

At a glance

  • Ingested and merged daily activity, sleep, and calorie tracking data.
  • Cleaned and filtered records, excluding days with missing data or zero activity.
  • Engineered features such as total active minutes and weekday labels.
  • Analyzed activity patterns by weekday and user behavior segments.
  • Built visual dashboards in Tableau to support wellness engagement insights.

Data & preparation

  • Raw datasets included: dailyActivity_merged.csv, sleepDay_merged.csv, and dailyCalories_merged.csv.
  • Each file was imported into R and cleaned using janitor::clean_names() to standardize column names.
  • Dates were parsed consistently across all datasets using lubridate::as_date().
  • Merged the datasets by user ID and date, producing a clean master dataset.
  • Removed records with missing data and zero steps to maintain quality.

Core import & merge (R):

library(tidyverse)
library(janitor)
library(lubridate)

daily_activity <- read_csv("dailyActivity_merged.csv") %>%
  clean_names() %>%
  mutate(activity_date = as.Date(activity_date, "%m/%d/%Y"))

daily_sleep <- read_csv("sleepDay_merged.csv") %>%
  clean_names() %>%
  mutate(sleep_day = as.Date(sleep_day, "%m/%d/%Y"))

daily_calories <- read_csv("dailyCalories_merged.csv") %>%
  clean_names() %>%
  mutate(activity_day = as.Date(activity_day, "%m/%d/%Y"))

merged_data <- daily_activity %>%
  inner_join(daily_sleep,   by = c("id", "activity_date" = "sleep_day")) %>%
  inner_join(daily_calories,by = c("id", "activity_date" = "activity_day")) %>%
  rename(date = activity_date)

Analysis & insights

  • Added calculated fields such as total_active_minutes by summing lightly, fairly, and very active minutes.
  • Derived day_of_week from the date field to analyze patterns by weekday.
  • Grouped data to explore averages by weekday and user-level activity trends.
  • Created Tableau dashboards to explore how activity, sleep, and calories interact.

Top behavioral insights:

  • Weekday vs weekend split: Most users were significantly more active Monday through Friday, with notable drops on weekends.
  • Sleep & activity correlation: Users with consistent sleep patterns tended to maintain higher levels of activity and calorie burn.
  • Active minute patterns: Days with high total active minutes were often aligned with peak calorie burn.

Visualizations

  • Used ggplot2 to build exploratory bar plots, histograms, and scatter plots for internal review.

  • Designed a Tableau dashboard to visualize:

    • Average steps by weekday
    • Calories burned vs. total active minutes
    • Sleep duration vs. daily steps

Recommendations

  • Encourage weekend activity through in-app challenges or reward prompts.
  • Nudge users with low sleep consistency to adopt wind-down routines — sleep quality links to engagement.
  • Create habit-forming nudges for users who show sporadic calorie/activity logs (e.g., auto-reminders).
  • Use clustering logic in future phases to segment users by wellness engagement style (e.g., active sleepers vs. calorie burners).

Reproduce

  1. Download the Fitbit public dataset from Kaggle.
  2. Load and clean files in R using tidyverse, lubridate, and janitor.
  3. Merge datasets by ID and date.
  4. Explore with ggplot2 and publish in Tableau.

Dashboard (optional link): Bellabeat Engagement Dashboard

Next

  • Segment by demographic if available (e.g., age group, gender).
  • Test correlation of steps/sleep/calories with clustering or regression.
  • Recommend personalized goal nudges based on user segment.

Updated: