Level-Up with Bellabeat

An analysis of consumer usage data in the fitness tracker industry targeting opportunities for growth.

This case study was the capstone project for my Google Data Analytics Professional Certificate. I had a choice of a few different studies but I found this one interesting because it spanned both business and healthcare data.

A Research Challenge

I liked the challenge because the dataset provided by Google did not meet all the standards for a good dataset following the ROCCC (Reliable, Original, Comprehensive, Current, and Cited) method of evaluating a dataset. The sample size (n= 30) was the bare minimum for acceptable, the duration of the study was limited, and the study was from 2016, so already five years old. It was ok but I decided I could also find a better, newer dataset; after all, fitness tracker usage must have changed a lot in 5 years!

Finding a Second Dataset

After much online digging I found a current study from 2021 with a much larger sample size, and a longer duration. Even better, the study followed college students and Bellabeat products are geared toward younger women. Most other studies followed older demographics. I decided to analyze the original given dataset first and use the new one for comparison. On to the analysis!

Company Overview

Bellabeat is a fitness tracker and fitness product company known for their attractive trackers that double as jewelry. The target customer is a young, fashionable, health-conscious woman.

The Business Task

Founder Urska Srsen wants the "marketing analytics team to...find opportunities for growth by analyzing smart device usage data to gain insight into how people are already using their smart devices. Then using this information, she would like high-level recommendations for how these trends can inform Bellabeat marketing strategy."

In one sentence: My task is to analyze current data and current trends, find opportunities for growth, and make high-level recommendations from a marketing perspective.

Current Trends in Fitness Trackers

• A simple interface

• Offering more granular data

• Gamification (addressed at the end of this case study)

A simple interface: Bellabeat already employs a simplified screenless tracker and a wealth of appealing related consumer products. The app is the connective element between all Bellabeat products. As such, the app must be both indispensable and a pleasure to use to inspire customers to purchase additional products and services.

Offering more granular data to its app --in the form of intensity of activity, and how that affects calorie burn-- would increase utility and engagement. Currently, Bellabeat offers the client data measured by:

• steps taken

• duration of activity

• heart rate

However, not all steps are equal. Steps burn calories at different rates, depending on the intensity of the exercise. The CDC’s “Physical Activity Guidelines for Americans recommend at least 150 minutes of moderate-intensity aerobic physical activity or 75 minutes of vigorous-intensity physical activity, or an equivalent combination each week.” (citation 1)

Comparing Calorie Burn at Different Workout Intensities

The reason the CDC specifies intensity in their above recommendations is the marked increase in calorie burn with increased workout intensity. We will look at that in the following 3 plots, created in R-Studio.

When heart rate is tracked, exercise intensity can be estimated. Using the given “Fitbit Intensity Set1” (citation 2), I plotted the different rates at which calories are burned based on intensity: “lightly active” (leisurely pace), “fairly active” (brisk walking), or “very active" (jogging or running).

Lightly Active minutes plotted against calorie burn:

In this dataset note that:

• calorie burn plateaus after roughly 175 lightly active minutes.

• It takes 175 minutes at this pace for the group (n=30) to burn 2,400 calories. So 80 calories per person in 3 hours.

The code:

```{r}

library(tidyverse)

library(readxl)

dailyAct <- read_excel("../input/fitbit-intensity-data/dailyActivity_lite.xlsx")```

```{r}

ggplot(data = dailyAct) + geom_smooth(mapping = aes(x = LightlyActiveMinutes, y = Calories))

```

Fairly Active Minutes Plotted Against Calorie Burn

In this dataset note that:

• calorie burn is most significant in the first 10 minutes

• It takes 10 minutes for this group moving at a brisk "fairly active" pace to burn 2400 calories. So 80 calories per person in about 10 minutes. That is a big time savings!

The code:

```{r}

ggplot(data = dailyAct) + geom_smooth(mapping = aes(x = FairlyActiveMinutes, y = Calories))

```

Very Active Minutes Plotted Against Calorie Burn

• This plot also shows a steep calorie burn in the first 10-12 minutes

• The calories burned in those first 12 minutes are slightly higher at 2500 or 83 calories/ person.

The code:

```{r}

ggplot(data = dailyAct) + geom_smooth(mapping = aes(x = VeryActiveMinutes, y = Calories))

```

The Takeaway from Dataset1

Just the first 10-12 minutes of moderate to vigorous activity burns significant calories. Minutes a day could lead to better health, especially if you increase the intensity.

Imagine You are the Customer

What if your Bellabeat app could give you more personal information about exactly how many minutes you need to spend on each level of activity?

What if it could save you time?

What if you could tailor it yourself?

And finally- What if it were fun?

A Larger Dataset and a Closer Look

Because the above was a small study of 30 participants over the course of two months, I brought in a larger dataset which had Fitbit data for hundreds of college-aged people over the course of almost two years.

I created the following three visualizations from this large-scale study by University of Notre Dame, Center for Network Science and Data (citation 3). The subset I used was data from the 63 most highly compliant participants who recorded >=350 days of data over the period of a year (2015-09-01 to 2016-08-31).

These can be seen together as a dashboard on Tableau Public: (https://public.tableau.com/views/FromGeneraltoGranularDataonExerciseIntensity/Dashboard?:language=en-US&:display_count=n&:origin=viz_share_link "From General to Granular Data on Exercise Intensity")

In this first chart, we see that as one might expect, most activity consists of Lightly Active Minutes (LAMs) as depicted by the light blue line. The distance between the light blue line and the gold line which shows Total Active Minutes (TAMs) represents remainder of activity from Fairly and Very Active Minutes combined.

The next chart shows total aggregate calorie burn was lowest in December. There was a sharp decline (17,634) in Very Active Minutes for the group which is mirrored by the decline in total calorie burn. In fact, this is only a decrease of 9 VAMs per subject per day. So, clawing back those 9 minutes of high intensity activity from the pressures of college midterms could go a long way toward maintaining health goals for these participants.

This second dataset is aligned with the the first dataset in showing that small increases (9-12 minutes/ day) in FAMs and VAMs can be significant.

The next chart gives us a personal look, comparing intensities between three similar subjects. Names are invented by me and all the data used was de-identified by the original study.

The Advantage of Granularity

Subtle Changes - Big Payoff

In the charts we see that Zara, Anna, and Mari all burn about the same number of calories per day. By comparing subjects with similar activity levels, we can better see how modest tweaks to one’s exercise intensity could yield better results in less time.

* Zara burns the most calories yet exercises the fewest hours per day. Should she become injured, such data could also help a very active person like Zara craft a gentler exercise routine that still meets her personal health goals.

* Anna exercises more Total Active Minutes (TAMs) than Zara but burns 377 fewer calories/ day. She could burn more calories by doubling her Fairly Active Minutes (FAMs) with brisk walks.

* Mari burns the fewest calories, yet she exercises for 1.2 hours more than Zara per day. By doubling her Very Active Minutes (VAMs) from 10 to 20, she could burn more calories and regain an hour of her day.

The Final Trend- Gamification

I did not forget that I promised to address gamification, the third and final trend I found in fitness tracker industry.

Gamification can complement new categories of data measurement, such as level of intensity, allowing the customer to set her own intensity challenges and be rewarded with badges, to “level up” and gain social-based rewards when milestones are reached. This links her interaction with the app to feelings of empowerment. In turn, that good feeling will help the user maintain engagement as she tackles the increasingly challenging personal fitness goals she sets.

A study of fitness tracker engagement by the National Institutes of Health (citation 4) (which surveyed 47 participants, 80.9% of whom were female) found the most used features in fitness trackers were:

* Rewards/Badges (59.6%)

* Notifications (52.2%)

* Challenges (42.6%)

High-Level Recommendations for Bellabeat:

* Use heart rate data to sort exercise into three levels of intensity

* Provide intensity metrics to customer in gamified & rewards-based way via the app

* Allow users to personalize their goals

Conclusion

With the promotion of valuable new intensity metrics gamified in the app, customers and Bellabeat will Level-Up -- together!

Citations

1. “How much physical activity do adults need?” Center for Disease Control and Prevention. U.S. Department of Health & Human Services. “Physical Activity Guidelines for Americans, 2nd edition.” Published 2018. (https://www.cdc.gov/physicalactivity/basics/adults/index.htm)

2. “FitBit Fitness Tracker Data. Pattern recognition with tracker data: Improve Your Overall Health.” By Mobius on Kaggle. (https://www.kaggle.com/arashnic/fitbit) Published 2020 Dec 16.

3. University of Notre Dame, Center for Network Science and Data “Using Fitbit data to examine factors that affect daily activity levels of college students” (https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0244747#sec009) Cheng Wang, Omar Lizardo, David S. Hachen. Published: 2021 Jan 6.

4. “The utility of wearable fitness trackers and implications for increased engagement: An exploratory, mixed methods observational study.” National Center for Biotechnology Information; National Library of Medicine; National Institutes of Health. By Zakkoyya H. Lewis, Lauren Pritting, Anton-Luigi Picazo, and Milagro JeanMarie-Tucker (https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6958644/) Published 2020 Jan 13.

5. “Gamified Wearable Fitness Tracker for Physical Activity: A Comprehensive Literature Review” By Inje Cho, Kyriaki Kaplanidou and Shintaro Sato (https://www.mdpi.com/2071-1050/13/13/7017) Published 2021 Jun 27.

Data Cleaning & Verification

1.) Fitbit Intensity Set1 (See second link in Citations)

• deleted 77 rows containing 0 steps

• corrected date format

2.) University of Notre Dame Data (See third link in Citations)

• Ensured each ID was 5 digit

• Ensured no whitespace

• Ensured single entry for each ID / date combination- removed 184 duplicates

• deleted rows containing null data for steps

• deleted rows containing null data for two or more of following: lightly active minutes, fairly active minutes, very active minutes.

• corrected date format

Tools Used:

R Programming with RStudio

Data Cleaning with Excel

Charts and Dashboard creation with Tableau

LinkedIn: Stella Bonaparte (https://www.linkedin.com/in/stella-bon)