03_share.Rmd

---
title: "Bellabeat Case Study — Visualizations"
author: "AHMAD DANIEL"
date: "2026-05-14"
output: html_document
---

---
title: "Bellabeat Case Study — Visualizations"
author: "Ahmad Daniel"
date: "`r Sys.Date()`"
output:
  html_document:
    theme: flatly
    toc: true
    toc_float: true
---

# Phase 5: Share

**Phase:** Share (Phase 5 of 6)

## Purpose

This notebook turns the findings from Phase 4 into seven polished 
visualizations. Each chart's title states an insight, not just a 
description, so the story is immediately legible to non-technical 
stakeholders.

## Setup

```{r setup, message=FALSE, warning=FALSE}
library(tidyverse)
library(lubridate)
library(scales)

# Load clean datasets
daily_activity   <- read_csv("../data/clean/daily_activity_clean.csv")
sleep_day        <- read_csv("../data/clean/sleep_day_clean.csv")
hourly_steps     <- read_csv("../data/clean/hourly_steps_clean.csv")
user_activity_summary <- read_csv("../data/clean/user_activity_summary.csv")
usage_frequency  <- read_csv("../data/clean/usage_frequency.csv")

# Re-apply factor levels
daily_activity$day_of_week <- factor(
  daily_activity$day_of_week,
  levels = c("Monday", "Tuesday", "Wednesday", "Thursday",
             "Friday", "Saturday", "Sunday")
)

user_activity_summary$user_type <- factor(
  user_activity_summary$user_type,
  levels = c("Sedentary", "Lightly Active", "Fairly Active", "Very Active")
)
```

## Bellabeat-inspired color palette

```{r palette}
# Soft, feminine wellness-brand colors
bellabeat_palette <- c(
  primary    = "#A8576E",   # dusty rose (primary brand color)
  secondary  = "#D9A6A0",   # blush
  accent     = "#6E8898",   # muted teal
  highlight  = "#E8B4B8",   # soft pink (for emphasis)
  neutral    = "#F4E8E1",   # warm cream
  dark       = "#3D3838"    # charcoal (text)
)

# Custom theme for visual consistency
theme_bellabeat <- function() {
  theme_minimal(base_size = 12) +
    theme(
      plot.title = element_text(face = "bold", size = 14, color = "#3D3838",
                                margin = margin(b = 8)),
      plot.subtitle = element_text(size = 11, color = "#6E6E6E",
                                   margin = margin(b = 12)),
      plot.caption = element_text(size = 9, color = "#9E9E9E", hjust = 0,
                                  margin = margin(t = 12)),
      axis.title = element_text(size = 11, color = "#3D3838"),
      axis.text = element_text(color = "#3D3838"),
      panel.grid.minor = element_blank(),
      panel.grid.major = element_line(color = "#EFEFEF"),
      legend.position = "top",
      legend.title = element_text(size = 10),
      plot.margin = margin(20, 20, 20, 20)
    )
}
```

---

## Chart 1: User segments — the majority is building habits, not chasing PRs

```{r chart-1-segments, fig.width=8, fig.height=5}
p1 <- user_activity_summary %>%
  count(user_type) %>%
  mutate(pct = round(n / sum(n) * 100, 1)) %>%
  ggplot(aes(x = user_type, y = n, fill = user_type)) +
  geom_col(width = 0.7) +
  geom_text(aes(label = paste0(n, " users\n(", pct, "%)")),
            vjust = -0.3, size = 4, color = "#3D3838") +
  scale_fill_manual(values = c(
    "Sedentary"      = "#A8576E",
    "Lightly Active" = "#D9A6A0",
    "Fairly Active"  = "#6E8898",
    "Very Active"    = "#E8B4B8"
  )) +
  scale_y_continuous(limits = c(0, 12), expand = c(0, 0)) +
  labs(
    title = "51% of users are Sedentary or Lightly Active",
    subtitle = "User type by average daily steps (n = 33)",
    caption = "Source: FitBit Fitness Tracker Data (Mobius, 2016)",
    x = NULL,
    y = "Number of Users"
  ) +
  theme_bellabeat() +
  theme(legend.position = "none")

p1
ggsave("../visualizations/01_user_segments.png", p1,
       width = 8, height = 5, dpi = 300)
```

**The story:** The Bellabeat customer is most likely a habit-builder, 
not a fitness elite. Marketing should reflect that reality.

---

## Chart 2: Sunday is the laziest day — 15% fewer steps than Saturday

```{r chart-2-day-of-week, fig.width=8, fig.height=5}
day_summary <- daily_activity %>%
  group_by(day_of_week) %>%
  summarise(avg_steps = round(mean(total_steps), 0))

p2 <- day_summary %>%
  ggplot(aes(x = day_of_week, y = avg_steps,
             fill = day_of_week == "Sunday")) +
  geom_col(width = 0.7) +
  geom_text(aes(label = comma(avg_steps)),
            vjust = -0.5, size = 3.8, color = "#3D3838") +
  geom_hline(yintercept = 10000, linetype = "dashed",
             color = "#9E9E9E", size = 0.5) +
  annotate("text", x = 1, y = 10300, label = "10K step target",
           color = "#9E9E9E", size = 3.2, hjust = 0) +
  scale_fill_manual(values = c("FALSE" = "#D9A6A0", "TRUE" = "#A8576E")) +
  scale_y_continuous(labels = comma, limits = c(0, 11000),
                     expand = c(0, 0)) +
  labs(
    title = "Sunday is the laziest day — 15% fewer steps than Saturday",
    subtitle = "Average daily steps by day of week",
    caption = "Source: FitBit Fitness Tracker Data (Mobius, 2016)",
    x = NULL,
    y = "Average Daily Steps"
  ) +
  theme_bellabeat() +
  theme(legend.position = "none")

p2
ggsave("../visualizations/02_day_of_week.png", p2,
       width = 8, height = 5, dpi = 300)
```

**The story:** Sunday's drop is a marketing opportunity. A Sunday 
motivation campaign would address the largest weekly behavior gap.

---

## Chart 3: Steps drive calories — but intensity matters too (r = 0.59)

```{r chart-3-scatter, fig.width=8, fig.height=5}
p3 <- daily_activity %>%
  filter(total_steps > 0) %>%   # remove non-wear days
  ggplot(aes(x = total_steps, y = calories)) +
  geom_point(alpha = 0.4, color = "#A8576E", size = 2) +
  geom_smooth(method = "lm", color = "#3D3838",
              fill = "#E8B4B8", se = TRUE) +
  scale_x_continuous(labels = comma) +
  scale_y_continuous(labels = comma) +
  annotate("text", x = 22000, y = 1500,
           label = "r = 0.59\n(moderate-to-strong)",
           color = "#3D3838", size = 4, hjust = 1, fontface = "italic") +
  labs(
    title = "More steps mean more calories — but only partially",
    subtitle = "Steps explain ~35% of calorie burn variance; intensity drives the rest",
    caption = "Source: FitBit Fitness Tracker Data (Mobius, 2016) | n = 863 days",
    x = "Total Daily Steps",
    y = "Calories Burned"
  ) +
  theme_bellabeat()

p3
ggsave("../visualizations/03_steps_vs_calories.png", p3,
       width = 8, height = 5, dpi = 300)
```

**The story:** Step count alone doesn't tell the whole calorie story. 
Bellabeat should promote intensity-based metrics — active minutes, 
heart rate zones — not just step counts.

---

## Chart 4: Activity peaks at 6 PM — the post-work commute window

```{r chart-4-hourly, fig.width=9, fig.height=5}
hourly_summary <- hourly_steps %>%
  group_by(hour) %>%
  summarise(avg_steps = round(mean(step_total), 0))

p4 <- hourly_summary %>%
  ggplot(aes(x = hour, y = avg_steps,
             fill = avg_steps >= 500)) +
  geom_col(width = 0.8) +
  geom_text(aes(label = avg_steps),
            vjust = -0.5, size = 3, color = "#3D3838") +
  scale_fill_manual(values = c("FALSE" = "#D9A6A0", "TRUE" = "#A8576E")) +
  scale_x_continuous(breaks = 0:23,
                     labels = c("12a", paste0(1:11, "a"), "12p",
                                paste0(1:11, "p"))) +
  scale_y_continuous(limits = c(0, 700), expand = c(0, 0)) +
  labs(
    title = "Activity peaks at 6 PM — the post-work window is gold",
    subtitle = "Average steps by hour of day; secondary peak at lunch (12-2 PM)",
    caption = "Source: FitBit Fitness Tracker Data (Mobius, 2016)",
    x = "Hour of Day",
    y = "Average Steps"
  ) +
  theme_bellabeat() +
  theme(legend.position = "none",
        axis.text.x = element_text(size = 9))

p4
ggsave("../visualizations/04_hourly_pattern.png", p4,
       width = 9, height = 5, dpi = 300)
```

**The story:** Notification timing should target 11 AM and 4-5 PM — 
right before users' natural activity peaks. Reminders after 8 PM are 
wasted on a low-activity audience.

---

## Chart 5: Users spend 39 minutes a night in bed but not asleep

```{r chart-5-sleep, fig.width=8, fig.height=5}
sleep_long <- tibble(
  metric = c("Time Asleep", "Restless (in bed, not asleep)"),
  minutes = c(419, 39)
) %>%
  mutate(metric = factor(metric,
                         levels = c("Restless (in bed, not asleep)",
                                    "Time Asleep")))

p5 <- sleep_long %>%
  ggplot(aes(x = "", y = minutes, fill = metric)) +
  geom_col(width = 0.4) +
  geom_text(aes(label = paste0(minutes, " min")),
            position = position_stack(vjust = 0.5),
            color = "white", size = 5, fontface = "bold") +
  coord_flip() +
  scale_fill_manual(values = c(
    "Time Asleep" = "#6E8898",
    "Restless (in bed, not asleep)" = "#A8576E"
  )) +
  labs(
    title = "Users spend 39 minutes a night in bed but not asleep",
    subtitle = "Average time asleep (419 min) vs. total time in bed (458 min)",
    caption = "Source: FitBit Fitness Tracker Data (Mobius, 2016) | n = 24 users",
    x = NULL,
    y = "Minutes per Night",
    fill = NULL
  ) +
  theme_bellabeat() +
  theme(legend.position = "top",
        axis.text.y = element_blank())

p5
ggsave("../visualizations/05_sleep_gap.png", p5,
       width = 8, height = 5, dpi = 300)
```

**The story:** 39 minutes nightly = 4.7 hours weekly of restless time 
in bed. Mindfulness features have a quantifiable problem to solve.

---

## Chart 6: The average day is 16.5 hours sedentary — even subtracting sleep, that's 9.5 hours awake and still

```{r chart-6-time-breakdown, fig.width=9, fig.height=5}
time_breakdown <- tibble(
  category = c("Sedentary", "Lightly Active",
               "Fairly Active", "Very Active"),
  hours = c(16.5, 3.2, 0.23, 0.36)
) %>%
  mutate(category = factor(category,
                           levels = c("Very Active", "Fairly Active",
                                      "Lightly Active", "Sedentary")),
         pct = round(hours / sum(hours) * 100, 1))

p6 <- time_breakdown %>%
  ggplot(aes(x = "", y = hours, fill = category)) +
  geom_col(width = 0.5) +
  geom_text(aes(label = ifelse(hours > 1,
                               paste0(hours, " hrs (", pct, "%)"),
                               "")),
            position = position_stack(vjust = 0.5),
            color = "white", size = 4.5, fontface = "bold") +
  coord_flip() +
  scale_fill_manual(values = c(
    "Sedentary"      = "#A8576E",
    "Lightly Active" = "#D9A6A0",
    "Fairly Active"  = "#6E8898",
    "Very Active"    = "#3D3838"
  )) +
  labs(
    title = "16.5 hours of every tracked day is sedentary",
    subtitle = "Only ~35 minutes is spent in fairly or very active states",
    caption = "Source: FitBit Fitness Tracker Data (Mobius, 2016)",
    x = NULL,
    y = "Hours per Day",
    fill = NULL
  ) +
  theme_bellabeat() +
  theme(legend.position = "top",
        axis.text.y = element_blank())

p6
ggsave("../visualizations/06_time_breakdown.png", p6,
       width = 9, height = 5, dpi = 300)
```

**The story:** The biggest behavioral opportunity isn't more workouts 
— it's reducing prolonged sedentary periods. Hourly stand reminders 
address the largest gap in the data.

---

## Chart 7: 88% of users tracked for 21+ days — churn isn't the problem

```{r chart-7-engagement, fig.width=8, fig.height=5}
engagement <- usage_frequency %>%
  count(usage_category) %>%
  mutate(pct = round(n / sum(n) * 100, 1),
         usage_category = factor(usage_category,
                                 levels = c("Low use (1-10 days)",
                                            "Moderate use (11-20 days)",
                                            "High use (21-31 days)")))

p7 <- engagement %>%
  ggplot(aes(x = usage_category, y = n, fill = usage_category)) +
  geom_col(width = 0.6) +
  geom_text(aes(label = paste0(n, " users (", pct, "%)")),
            vjust = -0.4, size = 4, color = "#3D3838") +
  scale_fill_manual(values = c(
    "Low use (1-10 days)"       = "#D9A6A0",
    "Moderate use (11-20 days)" = "#6E8898",
    "High use (21-31 days)"     = "#A8576E"
  )) +
  scale_y_continuous(limits = c(0, 33), expand = c(0, 0)) +
  labs(
    title = "88% of users tracked for 21+ days — the problem isn't churn",
    subtitle = "Once users adopt the device, they stick. Focus shifts to acquisition and depth.",
    caption = "Source: FitBit Fitness Tracker Data (Mobius, 2016) | n = 33 users",
    x = NULL,
    y = "Number of Users"
  ) +
  theme_bellabeat() +
  theme(legend.position = "none")

p7
ggsave("../visualizations/07_engagement.png", p7,
       width = 8, height = 5, dpi = 300)
```

**The story:** This contradicts the common smart device narrative. 
Bellabeat's marketing dollars should prioritize **acquisition** and 
**deepening engagement** — not retention.

---

## Summary

| # | Visualization | Headline Insight |
|---|---------------|------------------|
| 1 | User segments | 51% of users are in lower-activity tiers |
| 2 | Day of week | Sunday is the laziest day |
| 3 | Steps vs calories | Moderate correlation (r = 0.59) — intensity matters |
| 4 | Hourly pattern | 6 PM is the peak hour |
| 5 | Sleep gap | 39 restless minutes nightly |
| 6 | Time breakdown | 16.5 sedentary hours per day |
| 7 | Engagement | 88% tracked 21+ days — high commitment |

All charts exported to `visualizations/` at 300 DPI for high-quality 
display on GitHub and in presentations.

Proceed to **Phase 6: Act** in `docs/06_act.md` for recommendations.