Metricsnprose

BellaBeat Data Analysis Project Explained

Health and fitness should be a responsible and engaging journey. Bellabeat offers customer-focused fitness support, using data insights to enhance users’ approach while providing features that complement their goals.

About Bellabeat. 

Urška Sršen and Sando Mur founded BellaBeat, a high-tech company that manufactures health-focused smart products. Sršen used her background as an artist to develop beautifully designed technology that informs and inspires women. Collecting data on activity, sleep, stress, and reproductive health has allowed BellaBeat to empower women with knowledge about their health and habits. 

Since it was founded in 2013, BellaBeat has grown rapidly and quickly positioned itself as a tech-driven wellness company for women. By 2016, BellaBeat had opened multinational offices and launched multiple products. BellaBeat products became available through an increasing number of online retailers complimenting their e-commerce channel on their website. 

The company has invested in traditional advertising media, such as radio, out-of-home billboards, print, and television, but focuses on digital marketing extensively. BellaBeat invests year-round in Google Search, maintaining active Facebook and Instagram pages, and consistently engages consumers on Twitter. 

Additionally, BellaBeat runs video ads on YouTube and displays ads on the Google Display Network to support campaigns around key marketing dates. Sršen knows that analyzing  BellaBeat’s available consumer data would reveal more growth opportunities. 

She has asked the marketing analytics team to focus on a BellaBeat product and analyze smart device usage data to gain insight into how people already use their smart devices. Then, using this information, she would like high-level  recommendations for how these trends can inform BellaBeat’s marketing strategy

To satisfy Sršen’s data analysis needs, I have decided to follow a process that lets me gradually explore the data by asking the right questions, ensuring data integrity, and processing data to analyze the data properly to get Sršen actionable insights

Before we proceed to explore what Bellabeat is all about, let’s have a look at the scenario and my responsibilities as a data analyst to cater to this business scenario. 

Business Scenario

I am a junior data analyst working on the marketing analyst team at Bellabeat, a high-tech manufacturer of health-focused products for women. Bellabeat is a successful small company, but they have the potential to become a larger player in the global smart device market. 

Urška Sršen, co-founder and Chief Creative Officer of Bellabeat believes that analyzing smart device fitness data could help unlock new growth opportunities for the company. I have been asked to focus on one of BellaBeat’s products and analyze smart device data to gain insight into how consumers are using their smart devices.

The insights I discover will then help guide the marketing strategy for the company. I will present my analysis to the BellaBeat executive team along with high-level recommendations for BellaBeat’s marketing strategy. 

Characters In This Story. 

Characters

  • Urška Sršen: BellaBeat’s co-founder and Chief Creative Officer
  • Sando Mur: Mathematician and BellaBeat’s cofounder; a key member of the Bellabeat executive team
  • BellaBeat marketing analytics team: A team of data analysts responsible for collecting, analyzing, and reporting data that helps guide BellaBeat’s marketing strategy. 

I joined this team six months ago and have been busy learning about BellaBeat’s mission and business goals — as well as how you, as a junior data analyst, can help BellaBeat achieve them.

Products

  • BellaBeat app: The BellaBeat app provides users with health data related to their activity, sleep, stress, menstrual cycle, and mindfulness habits. This data can help users better understand their current habits and make healthy decisions. The BellaBeat app connects to their line of smart wellness products.
  • Leaf: BellaBeat’s classic wellness tracker can be worn as a bracelet, necklace, or clip. The Leaf Tracker connects to the BellaBeat app to track activity, sleep, and stress. 
  • Time: This wellness watch combines the timeless look of a classic timepiece with smart technology to track user activity, sleep, and stress. The Time watch connects to the BellaBeat app to provide you with insights into your daily wellness.
  • Spring: This is a water bottle that tracks daily water intake using smart technology to ensure that you are appropriately hydrated throughout the day. The Spring bottle connects to the BellaBeat app to track your hydration levels.
  • BellaBeat membership: BellaBeat also offers a subscription-based membership program for users. Membership gives users 24/7 access to fully personalized guidance on nutrition, activity, sleep, health, and beauty, and mindfulness based on their lifestyle and goals.

What My Plans For This Study Are. 

In this case, I choose to analyze data from FitBit users – a personal health tracker. My most important objective is to work with insights from FitBit secondary data to inspire BellaBeat with fresh, actionable business ideas/decisions. The result of my analysis should help guide BellaBeat’s marketing decisions especially for – Leaf (tracker bracelet) and Time (wellness watch). 

The leaf and time track and measures user wellness by integrating with the BellaBeat app. The app helps users make sense of the data they gather to provide insights into daily wellness, using attributes like sleep, weight, calories burnt, menstrual cycle, and mindfulness habits. 

In this case study, I will analyze FitBit data to find the most important problems BellaBeat might face and deliver recommendations on the tools to solve their business problems.The dataset for the project was provided via the Google Data Analytics Course in liasion with FitBit business tracker.  

The datasets for this project aren’t perfect but has some limitations. This project will not gloss over the limitations and make it seem like everything’s perfect. Instead, I will shed some light on possible ways to avoid these data errors in future. I expect to face several data issues during my career as a data analyst and this project will help me showcase how I intend solving these data issues. 

Below, you find me following the following steps: 

  • Develop a business scenario 
  • Understand the stakeholders that could be interested in my deliverables 
  • Pre-determine some guiding questions to help me satisfy my business demands 
  • Proceed with data analysis to find answers that satisfy the business recommendations 

Following this summary, each section will show how I walk through each phase of data analysis I will encounter in a work dat. 

I will also highlight the tools I have used in each process and how those tools will fit into delivering actionable business recommendations. 

Ask Phase 

Questions to guide my analysis include: 

  1. What are some trends in smart device usage?
  2. How could these trends apply to Bellabeat customers?
  3. How could these trends help influence BellaBeat’s marketing strategy?

To answer the above questions, I will have to report with the following deliverables: 

  1. A clear summary of the business task
  2. A description of all data sources used
  3. Documentation of any cleaning or manipulation of data
  4. A summary of your analysis
  5. Supporting visualizations and key findings
  6. Your top high-level content recommendations based on your analysis

Guiding questions

  • What is the problem you are trying to solve?

In this project, I am trying to help BellaBeat analyze data from FitBit to discover product loopholes they can exploit. When we discover these loopholes, we can deliver recommendations to help them improve customer user experience and in doing so, retain usage time on specific devices. 

  • How can your insights drive business decisions?

Insights from my analysis can help inform BellaBeat’s decisions on iterations that can make their products much more engaging. By improving product engagement, we can improve our customer base and make customers loyal. 

Key tasks

  1. Identify the business task
  2. Consider key stakeholders

They have been listed above already. 

Deliverable

A clear statement of the business task

Prepare Phase 

Before preparing the FitBit data for analysis, there are some inconsistencies that I feel need to be pointed out. Pointing out these data inconsistencies can help BellaBeat identify room for improvement and provide pointers for where future analysis may focus. 

Some inconsistencies I have uncovered with FitBit data include the following: 

  • The size of the data input is pretty limited as there are only inputs from 33 users. This tells us that the data is not comprehensive enough. 
  • Of the 33 volunteering users, only 8 users entered their weights, 12 users volunteered to enter their heart rate, and only 24 users entered sleep entries. 
  • Within the weighted dataset, some users did not enter information for all columns. – we will still work with all three datasets as they contain important information. 
  • The data is not firsthand – as it is from FitBit. This means that the available data may not completely align with Bellabeat’s user behavior. 
  • The data is not current with the collection date ranging from  4/12/2016 to 5/12/2016, which was about 7 years before the time of this case study. 
  • The data hasn’t been collected long enough. Since it is collected for only 30 days, the provided data may not account for all the variations in each user’s life/routine. 
  • Having only 33 users’ data, 30 entries each, will affect the reliability. Moreover, we are expecting 30×33=990 rows, however, there are 940 in the daily dataset.
  • Incomplete information could mean that some users either did not enter the information, were not wearing the tracker or the device did not collect the data properly. Also, some of the values were entered manually, for instance, that of the weight information. These and some other complications might result in biased data.

If these limitations were discovered in a real-life market analysis scenario, the limitations would need to be resolved before the analysis. Since this is a case study and I cannot control these data situations, I have to make do with what’s provided for comprehensive analysis. 

However in a real-life situation, here are some questions I would ask: 

  • Why do some users generate more data rows than others? Did FitBit use a device data collection system or did they turn them off?
  • Did users contribute data or at their convenience or were they told how often and when to use the app?
  • Are there measures that have been taken to remove sampling bias during data collection?
  • Is it possible to obtain the newer version of these datasets dataset?
  • Is it possible to obtain a similar dataset from BellaBeat for originality?

Before Data Cleaning, here is the metadata of every table in our database: 

Datasets Variables Num. of.Unique.Ids Num. of.Variables Num. of.Rows Missing.Values
dailyActivity_merged.csv Id, ActivityDate, TotalSteps, TotalDistance, TrackerDistance, LoggedActivitiesDistance, VeryActiveDistance, ModeratelyActiveDistance, LightActiveDistance, SedentaryActiveDistance, VeryActiveMinutes, FairlyActiveMinutes, LightlyActiveMinutes, SedentaryMinutes, Calories 33 15 940 0
heartrate_seconds_merged.csv Id, Time, Value 14 3 2483658 0
hourlyCalories_merged.csv Id, ActivityHour, Calories 33 3 22099 0
hourlyIntensities_merged.csv Id, ActivityHour, TotalIntensity, AverageIntensity 33 4 22099 0
hourlySteps_merged.csv Id, ActivityHour, StepTotal 33 3 22099 0
minuteCaloriesNarrow_merged.csv Id, ActivityMinute, Calories 33 3 1325580 0
minuteIntensitiesNarrow_merged.csv Id, ActivityMinute, Intensity 33 3 1325580 0
minuteMETsNarrow_merged.csv Id, ActivityMinute, METs 33 3 1325580 0
minuteSleep_merged.csv Id, date, value, logId 24 4 188521 0
minuteStepsNarrow_merged.csv Id, ActivityMinute, Steps 33 3 1325580 0
sleepDay_merged.csv Id, SleepDay, TotalSleepRecords, TotalMinutesAsleep, TotalTimeInBed 24 5 413 0
weightLogInfo_merged.csv Id, Date, WeightKg, WeightPounds, Fat, BMI, IsManualReport, LogId 8 8 67 65

 

Talk about the difficulties with formatting dates the right way and how we solved it with Power Query. 

Process Phase 

Data Cleaning and Processing 

During the cleaning process, we shall focus on ensuring data integrity and highlight some risks that may arise if we don’t maintain this integrity. Since we do not communicate directly with stakeholders in this hypothetical case study, I will try to solve as many integrity issues as possible.

For the case study analysis, the following columns have been selected: 

  1. dailyActivity_merged
  2. heartrate_seconds_merged
  3. hourlyCalories_merged
  4. hourlyIntensities_merged
  5. hourlySteps_merged
  6. minuteCaloriesNarrow_merged
  7. minuteIntensitiesNarrow_merged
  8. minuteMETsNarrow_merged
  9. minuteSleep_merged
  10. minuteStepsNarrow_merged
  11. sleepDay_merged
  12. WeightLogInfi_merged

 Below, I’ll provide a high-level description of the steps taken to ensure my data is clean and ready for analysis. The data cleaning steps are taken but not limited to what I’ve listed below: 

Alignment with Objectives (Merge)

This may include making necessary transformations removing the irrelevant information and organizing datasets in a suitable way for analysis. 

Looking at the datasets above, we can see a column_id. We can use the column_id to merge the datasets. All the datasets have a column for date but some have different header names for columns. For the columns with differing names, we can rename the said columns. Most of the dates also have date variables like date, hour, minutes, and seconds all in that single column. We shall split the column for date and time. Date is the date of observation and time includes (hours, minutes, and seconds). I’ll add columns for two usage groups as well. 

Dropping Columns 

Some datasets have columns that are irrelevant to what my analysis is trying to pursue. As with many analyses, some variables are not contextual to my analysis needs as well. Also, some data does not contain valid information. For example, the LoggedActivitiesDistance has a series of 0s.  Columns like this are irrelevant and need to be dropped from the datasets. 

  • Fat 
  • LogId
  • LogActivitiesDistance

Managing Outliers 

We have to ignore missing values and proceed with the analysis for the users who did not contribute data for the fields. There is also no way to fill in missing values but in the coming parts, we will recommend some ways Bellabeat can avoid these issues. 

Erase Duplicates 

We will be merging the datasets so that we can analyze daily, hourly, and minute data in an appropriate manner. In this step, it is expected that duplicate rows will be created when we merge them. We will simply remove those duplicates so that the results are accurate.

Validation

In the validation process, I need to make sure that the merged datasets follow the appropriate data types. In the validation process, I’d check if there were human errors in entering weight wrongly or if the times are consistent. Below is a display of the merged dataset based on daily, hourly, and minute entries. 

Analysis Phase

Most of the Analysis for this project was conducted with MySQL. For a complete breakdown of the steps I have taken to transform and analyze the data through MySQL, 

Lots of the data here are better visualized… Let’s import all of the necessary data into Power BI for in-depth analysis and visualization. 

  • I will first have to export the merged data from SQL and import it into SQL or connect the server via SQL directly. 
  • Since there are multiple datasets to analyze, I’d be better served directly connecting the server. 

Visualization Notes

Average calories for Fitbit users. 

Monday is the day with the most calories burnt – it kind of makes sense for Monday to have that amount of average calories burnt as it is the most active day of the week. Friday and Saturday also have a high calorie-burning average. 

  • This could mean that FitBit users are generally more active at the start, and at the end of the week.
  • The distribution of calories burnt looks right intuitively. 

Thursday reports the lowest average of calories burnt. 

Insert table for insight and interpretation. 

To probe further, let’s consider how activities are distributed daily. The chart will display the average calories per hour, steps, and heart rate. This will help us to compare those three variables and see if there is an interesting pattern at certain hours.

Average Steps .V. Active Distance with Calories Burnt as Bubble Size

From this visualization, an increased number of steps correlates with increased calories burnt and increased distance covered as well. 

This graph basically shows that our data is indeed accurate and we could do with other visualization to confirm our data’s authenticity. 

Calories Count Per Hour. 

17:00 appears to be the time with the highest calories burnt – After work hours? 

03:00 appears to be the time with the lowest calories burnt. 

Average Active Minutes By Week 

We can see that sedentary minutes have the highest count of average active minutes. The difference in total active minutes by weekday is also not entirely changing. This means that Bellabeat may want to invest in activity goals for users to meet as the users may already be trying to meet these goals themselves. 

They may already be trying to meet some personal goals each day and Bellabeat could encourage higher activity goals to increase daily activities minutes that are very active or fairly active. 

The activity level setup remains relatively the same all through with no noticeable changes over every day of the week. 

Average Steps By Weekday. 

Saturday is the day with the most steps, closely followed by Tuesday. Sunday has the least total steps – that may be explainable. 

After running the query, there wasn’t a whole lot of difference between each day in terms of average steps. With that said, Saturday had the highest average steps as well as the beginning of each week (Monday and Tuesday). 

We could potentially infer from this that the users wanted to be more active right after the weekend of rest (Sunday with the lowest total steps & Friday not too far behind) & that Saturday allowed for more time for activity & movement.

The Sum of Total Steps By Hour 

The top 5 hours recorded are: 

  • 17:00 – 195,109
  • 19:00 – 194,772 
  • 14:00 – 194,793
  • 10:00 – 190,463 
  • 18:00 – 187,994 

Deeper Look Into The Sleep Data. 

Average Of Total_Sleep_Minutes and Total_Time_In_Bed By Id

One user Id stands above all others while the clusters generally range from _ to _ 

The bottom two IDs are _ and _ 

The graphs show that most users who got at least 5 hrs. sleep had higher step counts. With this said, most users were not averaging the recommended 10,000 steps a day as noted in the Healthline article cited above.

Share Phase

Data Visualization In Power BI

The first step we took with imported data in Power BI starts with a brief summary to discover the distinction between users and get a general summary of what our data comprises of. I have used Cards and a Pie Chart to show distinctions / general components of FitBit data. 

As a reminder, the criteria for each user classification is: 

  • Active User – recorded activities for 21 – 31 days 
  • Moderate User – these user recorded activities for 11 – 20 days 
  • Light User – users that record their activities for 0 – 10 days 

The vast majority of users are active and light users are clearly the least class. We will be omitting the light users from further analysis as it may skew the observations and insights we give. 

How are activities distributed by weekday? 

Before we move to understand further user behavior, let’s consider the general activity distribution for users by weekday. 

Through this visual, we can see that Sedentary Minutes are the highest type of active minutes. What I noticed from the above graph is that the active minutes do not necessarily swing one way or another. It seems users are consistent in their active output each day. This information could mean that BellaBeat could leverage activity goals for users to meet as the users are already actively pursuing personal activity goals each day. Providing activities that align with the sedentary minutes can help improve what is already a favorable usage period for BellaBeat customers. 

Bellabeat could also encourage higher activity goals to increase daily activity minutes that are active or fairly active. 

Now that we have a rough idea of how active our users are, let’s break down other information from the devices to further understand of the users pattern of behaviour. 

Average Steps Per Weekday 

Rather surprisingly, FitBit user appear to take higher steps on Saturday but a expected, Sunday records the least number of average steps taken. If we remove the context from Saturday and Sunday, you’ll find that the reduction in average steps taken dwindles from the start of the week to its end (TMWFTS). 

We could potentially infer from this that the users wanted to be more active right after the weekend of rest (Sunday with lowest total steps & Friday not too far behind) & that Saturday allowed for more time for activity & movement. – To encourage further FitBit usage, we may look to provide activity suggestions that align with the increased average steps taken every year. 

With Saturday having the most average steps taken, it means FitBit users enjoy more freedom towards movement on Saturday. 

Saturday is the day with the most steps, closely followed by Tuesday. Sunday has the least total steps – that may be explainable. 

After running the query, there wasn’t a whole lot of difference between each day in terms of average steps. With that said, Saturday had the highest average steps as well as the beginning of each week (Monday and Tuesday). 

We could potentially infer from this that the users wanted to be more active right after the weekend of rest (Sunday with the lowest total steps & Friday not too far behind) & that Saturday allowed for more time for activity & movement.

Can We Infer further customer behaviors from both Active and Moderate Users? Let’s Find Out… 

For the Active Users, Tuesday is the only day they cross the CDC-recommended 8,000 steps. Saturday is the day when users barely reach the 8,000-step mark. The lowest week by steps is Sunday. 

For Moderate Users, Saturday is the day they properly cross the CDC-recommended 8,000-step mark. Wednesday is another day when moderate users cross the 8,000 steps mark. The lowest week by steps is Friday. 

If you notice the constant line across the chart, you’ll see the number – 8,000

Why 8,000? 

A Healthline.com article (“How many steps do I need a day?”) written by Sara Lindberg in 2019 cited a 2011 study by Tudor-Locke et. al. titled “How many steps/day are enough? for adults” which found that 10,000 steps/day is a reasonable target for healthy adults. 

I have already separated these users based on the number of steps taken before. 

What Can We Learn From These Charts? 

  • Saturday and Tuesday are favored days for both active and Moderate users. We can follow the initial advice of suggesting more FitBit-oriented activities on Saturday. 

How About Steps Taken Per Hour? 

Saturday – that has the highest number of steps taken, what hour of that day has the most steps taken? 

On Saturday, we’ll find that 13:00 to 18:00 are the hours that have the most steps taken. You’ll also find that this figure correlates with the line chart we can see below. 

FitBit will further its reach and customer usage by suggesting activities for users around this time frame. To explore changes to each day, please click this link (Link to Power Service Dashboard)

Removing Specific Date Selection From the Picture, what hours have the most steps taken?

The top 5 recorded steps were: 

  • 18:00:00 (6pm) – 187,994
  • 10:00:00 – 190,463
  • 19:00:00 (7pm) – 194,772
  • 14:00:00 (2pm) – 194,794
  • 17:00:00 (5pm) – 195,109

As suggested earlier, FitBit may want to suggest more activities – especially some sort of walking activities as users already walk a lot at these periods. 

Before we proceed further, let’s be sure that increased steps taken lead to more calories burnt. Since FitBit’s ethos is about keeping fit, burning calories is certainly an important aspect to consider especially as it can help us keep users motivated. 

As we would expect, an increased average calories burnt directly correlates with an increased average number of steps taken – the trendline points to that as well. It means FitBit can indeed showcase calories burnt notifications for users after each day (although this task is mainly a data science one, it can certainly be looked into). 

The average calorie burnt and distance covered follow the same trajectory as the prior TotalSteps data

The graph scatter chart shows that the more distance covered, the higher the average amount of calories burnt. 

For further data exploration, please click the link to the Power BI dashboard here – (Link to the Power BI dashboard here). 

To Reinforce this notion, see the scatter chart below: 

First, let’s have a look at the average minutes asleep across everyday of the week. 

Sunday has the highest average minutes spent asleep, followed by Wednesday. Despite users sleeping over 400 minutes on average, they are still under the 480 minutes of required sleep time. So FitBit needs to find a way for their users to sleep better – maybe including sleep notifications, sleep tracker, and time spent asleep – the highest average sleep time is still about 28 minutes below the required sleep range. 

This suggestion of improving healthy sleep habits among users begs the question – 

Is There An Increased Time In Bed Vs Time Asleep In Users?

To answer this question, let’s compare time asleep vs time in bed for the users. 

Plotting minutes asleep and time in bed against user ID shows that there is certainly room for improvement in most cases – FitBit can suggest midnight sleep tunes, meditation sounds, or general app features to improve the user’s sleeping habit. 

To explore how long each user sleeps in correspondence to the time they spend in bed, please refer to the Power BI app- here (Link to the Power BI dashboard here). 

Perhaps as expected, this graph shows a fairly consistent trend that the time spent in bed is not significantly different from the total sleep time, with the exception of a few users. This seems to suggest the accuracy of the device. Accurate sleeping monitoring can be attractive to customers who are seeking improvement and consistency in their sleep patterns.

There also seems to be an opportunity for Bellabeat to provide membership services that could help improve customers’ sleep patterns. One example could be introducing music or habits that will minimize the time they are awake in bed.

What if we separate and compare sleep data from the two main user groups – the active and moderate user groups? What can we find… 

While the sample size for the collected data and moderate users is quite small, it is an interesting observation that they cross the recommended 480 minutes in the constant line above. FitBit can find ways to include more sleep-inducing activities for moderate users as well in an attempt to convert their TimeInBed to TimeAsleep. 

Here we can visualize two separate variables; ‘Time Asleep’ and ‘Time in Bed’. Also within Figure 6, we have a constant line at 480 minutes. This is a reference line for the 8-hour recommended amount of sleep. From this, we can gather that all users who submitted data are below the recommended target for daily sleep time, with ‘low’ level users being the farthest away from the ideal.

Moreover, we see that users across all categories spend an average of 25 minutes longer in bed than they do asleep, indicating a delay between getting in bed and going to sleep, and waking up and getting out of bed.

Time Series Showing Average Calories per Hour of Each Day

  • Average calories burnt appear to peak between 17:00 and 19:00, on weekdays 
  • Thursday records a leveled spike in average calories burnt across multiple-hour zones. 10:00, 16:00, 19:00, and 21:00 all record significant spikes in the amount of calories burnt on Thursday. It doesn’t have a distinct, strong peak but is rather spread. 
  • Wednesday records the highest 17:00 peak in average calories burnt at 146. 

Relevant insight:

  • Insights about the peak hours indicate the activity time of users. They indicate that users are usually performing high intensity activities in the evening. 
  • Also Saturday the peak activities occur in the afternoon most likely because most users do not work on Saturday.
  • Thursday activities are much more spread than some of the other days. 
  • Saturday has a much higher pick than Friday but since activities start much earlier on Friday, we can consider Friday a more active day. 

Average Calorie Burnt By Fairly/Very Active Minutes 

Now, let’s look at sleep data. 

Average Calories By Weekday 

Moderate Users ( 3) have their most Very_Active_Minutes on Sunday. Thursday also records an overwhelming increase of Very_Active_Minutes. 

On Friday, Fairly_Active_Minutes are recorded way more than average active minutes 

The only Light User is active on Friday with Average_Fairly_Active_Minutes overwhelmingly higher than Average_Very_Active_Minutes. 

The Active Users generally record more average_fairly_active_minutes than average_very_active_minutes. 

Sunday represents a day with a more even distribution of fairly and very active minutes. 

Insight

This visualization illustrates the amount of daily ‘Very Active Minutes’ and ‘Fairly Active Minutes’ users undertake. This is segmented into two groups: ‘very’ and ‘fairly’ active level. We can see that the data for the active users doesn’t change drastically day-by-day like we can see with ‘medium’ level users. However, we also see that ‘medium’ level users have a clear preference for registering activity data on Saturdays.

The reference line on the graph is set at 22 minutes (amounting to 150 per week), which is the recommended amount of moderate physical activity an adult should undertake every day according to the CDC. We can see that all ‘high’ users surpass this amount, whereas for the ‘medium’ users it is sporadic, only going over 22 minutes for 4 out of the 7 days a week. However, both groups managed to achieve over 150 minutes per week across all days.

Compare – Average Total Minutes Asleep vs Average Total Steps

  • Users are most inactive on Sundays and Tuesdays 
  • Thursdays and Saturdays are the most active days for users 

Note that there is no specific order in how the calories descend from Saturday to Thursday. However, it is no surprise that Saturday users burn the most calories and Sunday most likely take some rest. The fact that Tuesday and Saturday are among the most active days also isn’t surprising. The bottom line is that this distribution of calories burnt looks right intuitively.

Now, we will create a visualization that shows how the activities are distributed per hour of each day. Figure 2 display not only the average calories per hour, but also steps and heart rate. This will help us to compare those three variables and see if there is an interesting pattern at certain hours.

Time Series Showing Average Calories Per Hour of Each Day

  • Average calories burnt appear to peak at between 17:00 and 19:00, on weekdays 
  • Thursday records a levelled spike in average calories burnt across multiple hour zones. 10:00, 16:00, 19:00, and 21:00 all record significant spikes in amount of calories burnt on Thursday. It doesn’t have a distinct, strong peak but is rather spread. 
  • Wednesday records the highest 17:00 peak in average calories burnt at 146. 

Relevant insight:

  • Insights about the peak hours indicates the activity time of users. They indicate that users are usually performing high intensity activities in the evening. 
  • Also Saturday the peak activities occur in the afternoon most likely because most users do not work on Saturday.
  • Thursday activities are much more spread than some of the other days. 
  • Saturday has a much higher pick than Friday but since activities start much earlier on Friday, we can consider Friday a more active day. 

Calories Burnt and Sedentary Minutes Relationship

Calories / Sedentary Minutes. 

On the other hand, shows a different trend line. This chart again confirms that the more users move and the more active they are, the more calories they burn.

Interpretation. 

There is a correlation between calories and one’s movements — the more steps taken, the longer distance moved, the longer active-time users had, the more calories are burnt. This is a strong data-driven insight that can be used for marketing strategy. This could help and encourage customers to be more active and move more as they can track and check their data and know that they are burning calories.

Act Phase 

BellaBeat’s female-centered approach paired with smart insights and body positivity has led to the development and embrace of wearable technology for women. From the available data and the visualizations we’ve had, it is clear users already interact with the product. 

This study aimed to harness data from the wearables to help customers unlock growth opportunities. Growth opportunities is basically users interacting with the product and improving their overall health. 

After combing through the data available, here are some things I found: 

  • Sedentary minutes are the highest type of active minutes. Users are pretty consistent in their active output daily. 
  • FitBit users appear to take higher number of steps on Saturdays. The increased number of steps taken dwindles from the start of the day to its end. 
  • Users want to be more active right after the weekend of rest. Saturday generally allows a lot more time for activity and movement as the users feel rested 
  • FitBit users barely cross CDC’s recommended 8,000 steps except on Saturday, Tuesday, and Wednesday – there can be room for growth here 
  • Saturday and Tuesday are clearly favored days for both active and moderate users. 
  • Hours 13:00 to 18:00 have the most steps taken on the days with the most steps taken. 
    •  18:00:00 (6pm) – 187,994
    • 10:00:00 – 190,463
    • 19:00:00 (7pm) – 194,772
    • 14:00:00 (2pm) – 194,794
    • 17:00:00 (5pm) – 195,109
  • Increased step taken correlates with an increased number of average steps taken. 
  • The more distance covered, the higher the amount of calories burnt. 
  • Sunday again has the highest average minutes spent asleep. Despite users crossing the 400-minute mark every day, they are still below the recommended 480 minutes. The highest average leep time is still around 28 minutes below the required sleep range. – there is room for improvement in the users sleep habits. 
  • Average calories burnt appear to peak at between 17:00 and 19:00, on weekdays 
  • Thursday records a levelled spike in average calories burnt across multiple hour zones. 10:00, 16:00, 19:00, and 21:00 all record significant spikes in amount of calories burnt on Thursday. It doesn’t have a distinct, strong peak but is rather spread. 
  • Wednesday records the highest 17:00 peak in average calories burnt at 146. 

Recommendation 

App Suggestions – Improving User Interaction. 

  • Through my analysis, I found that around 9% of users did not actively wear their FitBit devices while the other percent actively used the FitBit devices. I’d recommend a notification for users to recharge and wear their devices at the end of each day. 
  • My analysis also shows that most of the users do not reach the recommended 10,000 daily steps. Adding app notifications throughout the day or building self-goals may help users walk a little more and increase their total daily steps. 
  • Many people walk less than 7000 steps (46.5% of our sample). So I recommend to put an emphasis on this number. 10 thousand might feel too intimidating for some people, while 7 thousand steps sounds more reachable and according to the latest studies, it’s enough for health benefits.
  • Analysis also showed that most steps were taken in the evenings between 5pm – 9pm. This shows that the users are performing certain specific activities that warrant such increase in steps taken over those periods. It may be a great idea to develop app notifications that give user a heads up each time it seems like they aren’t moving at that specific period. 
  • There is no specific way for FitBit to find out what specific activities the users carry out in those periods – maybe I could suggest some. 
  • It may be of benefit to help users increase their sleep time by sending them notifications for them to unwind based on the average steps they take each day – Lastly, the data showed that users who averaged 5 hours of sleep or more also had a higher average step total.  
  • As Sundays tend to be less active for most of the people in our sample, some soft motivation could be helpful. By soft motivation I mean let people sleep and rest on this day (as they anyway will do), decreasing step goals but still sending few reminders to go out, take relaxing pleasant walks, and enjoy their Sunday.

Suggestions For FitBit Device 

On the basis of our analysis, we can offer some recommendations on marketing and product changes/updates Currently, most users track activities for over 21 days. To gain more actionable data, we need to find ways to entice higher usage levels with wear that processes an end goal. Users may feel discouraged from wearing or using everyday if they find the device or interface difficult to use, or the battery life of the smart device to be too short to wear for multiple days in a row.

To boost general activity, daily goals can be integrated into the device and app. These goals may be in line with the already achieved goals so users are immediately motivated to do more. Allowing users to flexibly set their goals based on their activity levels as it slowly rises. 

Finally, overall amounts of sleep are an issue for all users. Setting sleep reminders in app, and through the Leaf Fitness Tracker could assist users in reminding them when to go to sleep and when to wake up in order to meet their goals. Making the Leaf more comfortable to wear in bed, and further increasing battery life or speed of charging could entice more users into recording their sleep, as currently only 50% of FitBit users analysed do so regularly.

Helping users understand how everything the Bellabeat Leaf records throughout the day can assist them in leading healthier, more mindful, and more energetic lives will ultimately lead to customers who are satisfied with their purchases, and engage fully into the Bellabeat ecosystem of products.

Further Analysis with Additional Data

There were some limitations present in the data analysed. For a start, there were a few outlier observations that had to be excluded. Moreover, we worked with a relatively small sample size of only 33 users. With this in mind the analysis we were able to conduct was limited in scope. 

A future project could include not just a much larger number of participants, but also data pertaining to weight, height, age, as well as location data to help us dive deeper into trends across different demographics and regions. Allowing us to craft more bespoke recommendations targeting all of Bellabeat’s market segments.

Moreover, there is no way to distinguish aerobic and anaerobic activities in the data set. Recording such data would allow us to understand where activity is coming from, and how this impacts a user’s fitness level. Some may prefer cardiovascular exercise such as running or jogging which would lead to higher distances and steps taken, whereas other may stick to strength training, or even swimming, which would record significantly less steps but lead to similar results for a user’s overall fitness.

Marketing Recommendations

Again, while there were no demographic information to assist with any potential marketing recommendations there were some insights that might be helpful for the marketing team:

The data showed that 93.5% of users were ‘active users’ meaning that they utilized their FitBit consistently for 25-31 days throughout the data collection time frame. This high level of activity indicates that this group of users are invested in utilizing their fitness tracker. I would recommend marketing Bellabeats products to customers who may already own a tracker or are already invested in wellness or learning about their health. Showcasing this product as woman-focused as its unique features may convience customers who already own a wearable to switch to one of Bellabeat’s products for the benefit of more targeted insights.

The data also showed (as mentioned above) that the total steps tracked were highest during the lunch time frame and the 5pm – 7pm time frame. This indicates that most users have a set routine — usually fitting in the most activity during lunch or potentially after work. I’d recommend that Bellabeat market their products to customers most likely living around this particular type of routine — customers with a set job schedule and parents with a set daily routine.

The key message for the Bellabeat online campaign

The Bellabeat app is not just another fitness activity app. It’s a guide (a friend) who empowers women to balance full personal and professional life and healthy habits and routines by educating and motivating them through daily app recommendations.

Ideas for the Bellabeat app

  • Average total steps per day are 7638 which a little bit less for having health benefits for according to the CDC research. They found that taking 8,000 steps per day was associated with a 51% lower risk for all-cause mortality (or death from all causes). Taking 12,000 steps per day was associated with a 65% lower risk compared with taking 4,000 steps. Bellabeat can encourage people to take at least 8 000 explaining the benefits for their health.
  • If users want to lose weight, it’s probably a good idea to control daily calorie consumption. Bellabeat can suggest some ideas for low-calorie lunch and dinner.
  • If users want to improve their sleep, Bellabeat should consider using app notifications to go to bed.
  • Most activity happens between 5 pm and 7 pm – I suppose, that people go to a gym or for a walk after finishing work. Bellabeat can use this time to remind and motivate users to go for a run or walk.
  • As an idea: if users want to improve their sleep, the Bellabeat app can recommend reducing sedentary time.

Discussion 

In this section, I will discuss the most important findings and suggestions from this case study. While working on this case study I realized that I could build much stronger analysis and extract more information. 

However, since the data integrity was violated, it was not worth doing that. With only 33 users and 30-day user data, we wouldn’t be able to get the big picture and produce accurate results. There are important variables such as weather within each season or variability of user characteristics, which can not be ignored. After all, the fact that the data is bad makes results of any analysis unreliable.

Perhaps, one of the most important suggestions here is that, as mentioned before, Bellabeat should focus on integrating data science into its recommendation system. We need to refer to market research, but it can be assumed that what users want is probably an app that helps them track their progress accurately and help them reach their goals effectively. Data scientists can develop classification algorithms such as neural networks that will accurately predict the type of activity at a given time.

 

Leave a Comment

Your email address will not be published. Required fields are marked *

Scroll to Top