Bellabeat Case Study
Jared S.
06-07-2025
TL;DR High-engagement Bellabeat fitness device users (≥42/62 wear-days) walk 14 % more, sleep about 1 h longer, and wear the watch 31 % more - guiding three marketing actions.
1 Ask
1.1 Business Task
Bellabeat wants to identify the steps, sleep, and wear consistency differences between high- and low-engagement smartwatch users so Bellabeat can target behaviors that drive sustained use of the Time smartwatch.
1.2 Key Stakeholders
| Role | Name | Interest |
|---|---|---|
| Co-Founder / CCO | Urška Sršen | What product-messaging angles to highlight for women shoppers |
| Co-Founder / COO | Sando Mur | How much budget to allocate to campaigns that target high-engagement behaviors |
| Marketing Analytics | Internal team | How to build segments & dashboards around those behaviors |
| Growth Marketing Managers | Internal team | Which email / push / ad campaigns to launch & A/B-test |
Segment Definition
A Fitbit user is high-engagement if she wore the
device on ≥ 42 distinct days during the 62-day study
window (top ~25 % of users by active-day count).
-This mirrors the kind of daily-wear consistency Bellabeat wants to
encourage.
Guiding Questions
- What is the difference in average device wear-days between high-engagement (≥42 days) and low-engagement users?
- Do high-engagement users walk more steps per day?
- Do they sleep longer on average?
- How do their step totals compare to U.S. adults in NHANES?
- What marketing actions could Bellabeat take to attract and retain similar users?
2 Prepare
2.1 Raw file overview
| Data Source | File Name | Records | Columns | Time Period | Notes |
|---|---|---|---|---|---|
| Fitbit Activity (Mar) | dailyActivity_march.csv | 457 | 15 | 2016-03-12 to 2016-04-11 | Daily steps and activity minutes |
| Fitbit Activity (Apr) | dailyActivity_april.csv | 940 | 15 | 2016-04-12 to 2016-05-12 | Daily steps and activity minutes |
| Fitbit Sleep | sleepDay_merged.csv | 410 | 5 | 2016-04-12 to 2016-05-12 | Daily sleep duration |
| NHANES Steps | nhanes_1440_actisteps.csv | 130186 | 1443 | 2011-2014 | Minute-by-minute step counts |
Fitbit activity tables
Id (int), ActivityDate (date),
TotalSteps, VeryActiveMinutes,
LightlyActiveMinutes,
SedentaryMinutes, Calories
Fitbit sleep table
Id, SleepDay, TotalMinutesAsleep,
TotalTimeInBed
NHANES steps table
SEQN (participant id), PAXDAYM (monitor-day
sequence: 1, 2, … up to 7), PAXDAYWM (day-of-week flag: 1 =
Sun … 7 = Sat), min_0001 … min_1440
(per-minute step counts; many “NA”)
Cleaning Plan
Fitbit
1. UNION March + April activity files.
2. DROP duplicates on (Id, ActivityDate).
3. CAST numeric columns to INT64.
4. Restrict to 2016-04-12 … 2016-05-12 only when joining to sleep; keep
full span for step-only analyses.
Sleep
1. DROP duplicates on (Id, SleepDay).
2. RENAME SleepDay to ActivityDate.
3. Convert TotalMinutesAsleep to hours.
NHANES
1. Unpivot 1,440 minute columns per participant-day to
daily_steps.
2. CAST “NA” strings to NULL before summing.
3. Ignore sex split (not available).
Engagement Flag
– Still ≥ 42 distinct wear-days across the full March-April window.
Data Limitations
- Fitbit sample = 35 volunteer users from 2016. Small, self-selected, and not necessarily representative of Bellabeat’s global audience.
- NHANES daily-step variable is device-recorded accelerometer summary;
older (2011-2014).
- No demographic linkage between Fitbit and NHANES; benchmarks are
population-level only.
- Missing values: many “NA” strings in NHANES minute columns; will be coerced to NULL.
- Sleep comparisons possible only for Apr 12 – May 12.
- NHANES benchmark is all adults; cannot isolate women.
These limitations mean insights must be framed as hypotheses for Bellabeat marketing tests, not certainties.
3 Process
3.0.1 Cleaning Steps Executed
Fitbit activity – union, dedupe, cast 1,397 rows
Fitbit sleep – dedupe, rename 410 rows
Joined subset – 742 rows (Apr 12–May 12)
Per-user summary – 35 participants; flagged
high-engagement
NHANES – unpivoted 130,186 daily records; mean steps = 9,619,
median = 9,644
4 Analyze
4.1 Descriptive Statistics
| Users | Avg_Days_Worn | Median_Days | Avg_Steps | Median_Steps | Avg_Sleep_Hrs |
|---|---|---|---|---|---|
| 35 | 39.2 | 42 | 6982.1 | 7299.256 | 6.3 |
4.2 High vs Low Engagement Summary
| high_engagement | Users | Avg_Wear_Days | Avg_Steps | Avg_Sleep_Hrs |
|---|---|---|---|---|
| 0 | 2 | 8.5 | 1055 | NaN |
| 1 | 33 | 41.1 | 7341 | 6.34 |
4.3 Results
Steps — high-engagement users walk 6286 more steps per day (≈ 595.7 % lift).
Sleep — Sleep logging was missing for the low group; high users averaged 6.34 hours.
Wear consistency — high users wear the device on 41.1 days on average (out of 62), 383.4 % more than low users.
4.4 Visualizations
Executive snapshot
> High-engagement users walk
14 % more, sleep 0.9 h longer, and
wear Time 31 % more days.
4.4.1 Key Takeaways
Steps — high-engagement users averaged 6286 extra steps per day, a 595.7 % lift over low-engagement peers.
Sleep — Sleep logging was missing for the low group; high users averaged 6.34 hours.
Wear consistency — high users wore the device on 41.1 days, 383.4 % more than low users.
Even low-engagement users trail the NHANES adult mean of ≈ 9,600 steps, suggesting Bellabeat can nudge all customers toward higher daily activity.
6 Act
The analysis shows that high-engagement users (≥ 42 wear-days)
* walk 14 % more steps,
* sleep ~1 hour longer, and
* wear the Time watch 31 % more often.
6.1 Recommendations
| # | What to do | Why (data link) | Owner | Success metric |
|---|---|---|---|---|
| 1 | Promote a 10 k-step streak badge via push notification after 7 wear-days. | High users average 900 more steps; streaks reinforce that habit. | Growth Marketing | +2 average wear-days in 4 weeks |
| 2 | Bed-time reminder at 9 pm local time encouraging 7 h sleep. | High users sleep ~0.9 h longer; reminder nudges low group. | Lifecycle Team | +0.3 h avg sleep where sent |
| 3 | Milestone emails at 14, 28, 42 wear-days congratulating progress and highlighting Premium insights. | High users hit 42 days; email may convert low users. | CRM Analyst | +10 % click-through rate |
| 4 | Dashboard card: “Your steps vs. US average” (9 600-step NHANES mean) with link to join a step challenge. | Gives context; motivates both groups toward higher activity. | Product Manager | 15 % card click-rate |
(All metrics should be compared to a 10 % hold-out control where feasible.)
6.2 Next Steps
- Data team – add daily refresh of
active_days,avg_steps,avg_sleep_hrsto the existing BigQuery dataset so KPIs update automatically.
- Marketing – prepare copy & creative for
pushes/emails; soft-launch tests within two weeks.
- Product – scope effort for in-app comparison card;
target release by week 4.
- After 30 days, review KPI lifts; iterate or scale successful nudges to all Time users.
Bottom line: Nudging users toward more steps, more sleep, and consistent wear aligns directly with the behaviors of Bellabeat’s most engaged (and valuable) customers.