Analysis of Smart Device Usage for Bellabeat Marketing Strategy
By Dane Tipene
June 18, 2024
Bellabeat, founded by Urska
Srsen and Sando Mur, specializes in health-focused smart products
designed to inform and inspire women. Since its inception in 2013,
Bellabeat has rapidly grown and positioned itself as a leader in
tech-driven wellness products for women.
Phase 1: Ask
Business Task
Analyze non-Bellabeat smart device usage data to gain insights that will enhance Bellabeat’s marketing strategy. Apply these insights to a selected Bellabeat product for targeted improvements.
Key Stakeholders
- Urska Srsen: Bellabeat’s cofounder and Chief Creative Officer
- Sando Mur: Bellabeat’s cofounder and key member of executive team
- Bellabeat Marketing Team
Market Insights
According to Fortune Business Insights, as well as studies and statistics from ValuePenguin (April 2022) and RunRepeat (October 2021), the following insights were observed:
- Market Growth: The global fitness tracker market is projected to grow from USD 53.94 billion in 2023 to USD 290.85 billion by 2032, at a CAGR of 21.3%.
- User Engagement: 92% of smartwatch wearers use their devices for health and fitness monitoring, with 88% claiming they helped in achieving fitness goals. Smartwatches hold the largest smart device type market share.
- Application: Running is the most common use, with 42.8% of users globally prioritizing this feature.
- Shipping Volume: 445 million wearable fitness devices were shipped, highlighting strong demand and market potential.
- Pandemic Impact: Fitness tracker shipments and revenue grew over 31% during the COVID-19 pandemic, indicating increased consumer interest in health.
- Gender: Women are nearly 40% more likely to use fitness trackers than men, benefiting female-focused brands like Bellabeat.
- For more detailed insights, you can refer to the full article here.
Phase 2: Prepare
Data Source
- Fitabase dataset made available through Kaggle user Mobius and licensed under CC0: Public Domain
- Dataset generated from 33 eligible Fitbit user logs from smart devices
- Acknowledgements - Robert Furger, Julia Brinton, Michael Keating, Alexa Ortiz
Data Credibility
The data is reliable but dated, originating from 2016 and collected over 31 days from 33 participants.
The small sample size of 33 participants reduces the confidence level of the derived insights. Additionally, the absence of demographic information such as gender, age, and location makes it highly unlikely to determine if the data accurately represents the population. This also hinders the identification of female-specific trends.
Chosen Tools
I initially examined the datasets in Excel to sort and filter the data. However, the ‘Minute’ datasets contained millions of rows, making it impractical to manage in Excel. Therefore, I loaded all the data into R Studio for more efficient cleaning and analysis.
- Excel
- R
- R Packages
Chosen Datasets
- dailyActivity_merged
- dailyCalories_merged
- dailyIntensities_merged
- dailySteps_merged
- heartrate_seconds_merged
- hourlyCalories_merged
- hourlyIntensities_merged
- hourlySteps_merged
- minuteintensitiesNarrow_merged
- minutesSleep_merged
- sleepDay_merged
- weightLogInfo_merged
Phase 3: Process
Cleaning Log
Inspect and Preview All Datasets
- Inspected Data using ‘str()’ function
- Searched for matching columns for potential merging, and checked for inconsistencies in column names and formats
- Found “Id”, “ActivityDate”, and “ActivityHour” columns suitable for merging separate daily and hourly datasets
Dataset Transformations
- Cleaned inconsistent column names to match appropriate datasets
- Renamed date/time columns to either “ActivityDate”, “ActivityMinutes”, or “ActivitySeconds”
- Renamed ‘daily_activity’/“TotalSteps” to “StepTotal” to align with other ‘steps’ datasets
- Standardized inconsistent date/time formats
- A total of 33 participants were identified, which is 3 more than originally stated in the dataset description.
- Identified 65 missing values in the ‘weight_log’ dataset and chose not to delete these rows, as it would remove 99% of the data
- Identified and removed 3 duplicates from the ‘sleep_day’ dataset
- Identified and removed 543 duplicates from the ‘minute_sleep’ dataset
- Found the ‘minute_sleep’ dataset to be of no value and excluded it from the analysis
- Merged Daily datasets together
- Merged Hourly datasets together
- Reduced datasets from 12 to 5: two merged datasets (‘daily_combined’ and ‘hourly_combined’) and three separate datasets (‘minute_intensities’, ‘heartrate_seconds’ and ‘weight_log’)
Phase 4: Analysis and Insights
User Engagement Patterns
- The data measures four primary metrics: Physical Activity, Sleep,
Heart Rate, and Weight.
- Among these metrics, users clearly valued tracking their physical activity the most, while weight and heart rate were tracked the least. This could be an opportunity to investigate why these metrics are less tracked and how to encourage more users.
- The high percentage of “Consistent Use” participants suggests that the device/platform is highly effective in engaging users regularly. This provides a solid foundation for both marketing the product and developing new features.
- At the same time, the presence of “Moderate Use” and “Minimal Use” users indicates an opportunity to further refine and personalize the user experience to increase engagement across the board.
## # A tibble: 1 × 4
## AverageDays MinDays MaxDays PercentageOfTotalDays
## <dbl> <int> <int> <dbl>
## 1 28.5 4 31 91.9- The average daily usage of their smart devices was measured at 28.5 days out of 31, indicating that participants used their devices 91.9% of the time.
- Only one user used their device for 4 days, while the next closest participant used their device for 16 days.
Activity Level Categories
- Fitabase categorizes activity levels into four groups: Sedentary,
Lightly Active, Fairly Active, and Very Active.
- According to this LiveStrong
article, the “10,000 Steps Project” suggests five levels of activity. I
combined the last two levels into the Very Active category (10,000 steps
and above). By averaging the Daily Total Steps measured for each
participant, I was able to categorize each participant into their
respective Activity Level.
Even Distribution
- The distribution across categories is relatively even, with no single category dominating the user base.
- This suggests that users have varied levels of activity, and the
platform caters to a diverse group of individuals with different
activity levels.
Middle Range Dominance
- The middle categories (Lightly Active and Fairly Active) each account for 27.3% of users.
- This indicates that most users fall into moderate activity levels,
which could be targeted for further engagement and
improvement.
Opportunities for Engagement
- Users in the Sedentary category could be targeted with personalized interventions and encouragement to increase their activity levels. Reward systems could be utilized to encourage movement.
- Users in the Very Active category might be interested in advanced features or challenges to maintain their high activity levels.
Time Distribution Across Activity Levels
- The majority of each day is spent at the sedentary level, accounting for 84.1% or 19 hours and 49 minutes.
- This is approximately composed of 7 hours in bed, likely 8-10 hours of working and commuting, and possibly an hour or two of watching TV.
- Lightly Active time accounts for 14% or 3 hours and 25 minutes, likely spent on routine daily tasks.
- Very Active time is 1.97% or 28 minutes, potentially spent on vigorous exercise.
- Lastly, 1.43%, or 20 minutes, is spent in the Fairly Active category, which could also involve exercise.
## # A tibble: 4 × 4
## ActivityLevel TotalAvgMinutesSpent TotalAvgPercentage TotalAvgHoursSpent
## <chr> <dbl> <dbl> <dbl>
## 1 FairlyActive 20.4 1.43 0.339
## 2 LightlyActive 206. 14.4 3.43
## 3 Sedentary 1190. 84.1 19.8
## 4 VeryActive 28.3 1.97 0.471
Activity Patterns
Peak Activity Times
- High activity times range between 12 PM and 3 PM, with peak activity occurring between 5 PM and 7 PM.
- Peak activity is most likely due to post-work workouts.
- Send reminders or tips for physical activities during high activity times (12 PM to 3 PM) and peak times (5 PM to 7 PM). Additionally, launch fitness challenges or competitions that align with these peak activity periods to encourage participation and engagement.
Average Steps, Distance, Calories
## # A tibble: 1 × 3
## OverallAvgStepsPerDay OverallAvgDistancePerDay OverallAvgCaloriesPerDay
## <dbl> <dbl> <dbl>
## 1 7527. 5.41 2283.Steps
- The overall average steps fall almost exactly between the lightly active and fairly active categories at 7,519.27. This is close to the commonly recommended 10,000 steps per day, suggesting that users are relatively active but could still improve to reach the recommended goal through alerts, step challenges, encouragement, or reward systems.
- Identify users with significantly lower step counts and provide them with personalized interventions or encouragement to increase their activity levels.
- Segment users based on their average step counts to tailor content and features. For instance, beginners/sedentary users may need basic guidance and motivation, while more active users might benefit from advanced analytics and challenges.
Distance
- Unable to completely determine if distance is measured in km or M however, according to the OmniCalculator site, the distance is likely measured in km.
- In alignment with step counts, the average distance traveled could be slightly increased to healthier levels through encouragement features.
Calories
- The average daily caloric expenditure for each user from the Fitabase data is 2,282 calories, which is slightly below the average caloric burn for a moderately active person (2,437.5 calories) as referenced by WebMD. This suggests that, participants in the Fitabase data may be slightly less active than the general moderately active population.
- There is potential to encourage users to increase their activity levels slightly to meet or exceed the average caloric expenditure of a moderately active person. Note: This dataset is not gender-specific. Since Bellabeat is a female-focused brand, it is important to consider the average caloric expenditure of a moderately active female when utilizing this information for marketing and product improvements.
Calories vs Steps
- As expected, there is a strong positive correlation showing that as users’ physical activity (measured by steps) increases, their caloric expenditure also rises. This underscores the importance of physical activity in managing energy balance and weight.
Daily Activity Stats
- Tuesday (8125 steps) and Saturday (8153 steps) have the highest average steps, both falling into the “FairlyActive” category.
- The weekend days (Saturday and Sunday) show a contrast in activity levels. Saturday is one of the most active days, while Sunday is one of the least active days.
- Sunday has the lowest average steps (6933), falling into the “LightlyActive” category.
- This could be an opportunity to develop, promote, and market weekend specific fitness plans or programs that cater to both high activity days like Saturday and low activity days like Sunday.
- Highlight the benefits of staying active over the weekend by offering specific workout challenges and promoting family-focused active activities. This could broaden the appeal, encouraging users to involve their family members, potentially increasing the number of new users and enhancing overall user satisfaction and loyalty.
- AAP = Average Active Person Caloric Burn (2437.5)
- AAF = Average Active Female Caloric Burn (2150)
- Since demographic data is nonexistent, we must assume these results represent a mixed-gender population. However, setting a daily caloric target based on gender and age could help motivate users more effectively.
- The average calories burned range from 2199.571 to 2356.013, showing a relatively consistent caloric burn across the week.
Sleep Patterns
Sleep Metrics
- Two metrics were recorded: TotalTimeAsleep and TotalTimeInBed. Both were converted into hours.
- On average, users spent 7 hours in bed and 6hrs 19 mins asleep. This falls 41 minutes short of the recommended sleep time for adults, as advised by the Sleep Foundation.
- This presents an opportunity to promote healthier sleep patterns by providing content and tools to help users manage their time better and develop relaxing end-of-day routines. Features can include bedtime reminders, relaxation techniques, and advice on minimizing screen time before bed.
- Currently, 72.7% of all participants utilize the sleep recording feature.
- Among these users, only 50% can be categorized as consistent users of this feature, with 12.5% using it moderately and over a third (37.5%) using it minimally.
- On average, users recorded their sleep data only 17 out of 31 days, or 55% of the time.
- Less than desirable usage could be due to charging device overnight. Further exploration of this is needed.
Sleep Quality Vs Physical Activity
- Interestingly, there is a slight negative correlation between physical exertion and sleep duration, which contradicts the advice from the Sleep Foundation.
- However, this could be due to other factors not recorded in the dataset, including diet, screen time, TV habits, water intake, stress, and work/life commitments.
Heart Rate
- Only 14 users (42.4%) utilized this feature.
- The average bpm of 79.98 falls within the normal resting heart rate range (60-100), which is good.
- The maximum heart rate recorded was 203, which is within the expected range. A commonly used formula to estimate your maximum heart rate is 220 minus your age.
- The lowest heart rate recorded was 36, which is lower than that of a well-trained athlete (40 bpm).
- All general healthy heart rate ranges referenced from the Mayo Clinic, and Heart Online.
Weight Insights
- This feature is utilized by the smallest group of users, accounting for 24.2% or 8 people.
- Those who used this feature fall into a diverse weight range, from 52.6 kg (116 lbs) to 134 kg (294 lbs).
- The average weight is 77.8 kg (171.54 lbs), and the average BMI is 28.
- Since the average BMI of 28 falls into the overweight category, as referenced by the CDC, providing gentle support towards achieving a healthier range is recommended.
- The BMI values range from 21.5 to 47.5. Creating and providing segmented advice based on BMI categories could significantly enhance user engagement and the utilization of this feature.
Conclusion
To drive business through marketing campaigns, leverage insights to
enhance user engagement and health tracking features. Focus on promoting
personalized physical activity and sleep plans, and integrate additional
health monitoring tools. Emphasize the value of tailored content and
motivational strategies to support users’ health goals while balancing
the frequency of reminders to avoid annoyance. Develop and highlight new
products like a Bluetooth Smart Scale and emphasize user testimonials in
campaigns. This strategic approach will increase user satisfaction and
feature utilization, fostering a health-conscious user base and driving
growth.