Bellabeat Analysis Project: Utilizing Trends in Health Tech Device Usage to Guide the Marketing Strategy
- 1. Bellabeat Case Study: Utilizing Trends in Health Tech Device Usage to Guide the Marketing Strategy Hiba Shawa | November 7th 2022
- 2. Agenda 1. Project overview and Purpose 2. Data and analysis 3. Fitbit: Questions Asked and Answered 4. What we’ve learned, possible questions, and recommendations
- 3. Project: Overview and Purpose • Bellabeat is a high-tech manufacturer of health- focused products for women. • One of it’s products, Time, is a wellness watch that combines the timeless look of a classic timepiece with smart technology. • “Time” tracks user activity, sleep, and stress. • The Time watch connects to the Bellabeat app to provide users with insights into their daily wellness. • Marketing is a key pillar in the growth of this company. • In this project I will be analyzing a similar health tracking watch to find trends… The aim of this project was to determine how consumers use non-Bellabeats smart devices (ie. FitBit activity trackers) and use these insights to drive its marketing strategy.
- 4. An overview of Fitbit data Fitbit Fitness Tracker Data was obtained from Kaggle.com Title: “Pattern recognition with tracker data: : Improve Your Overall Health” https://www.kaggle.com/datasets/arashnic/fitbit About dataset - From Kaggle.com: • “This dataset generated by respondents to a distributed survey via Amazon Mechanical Turk between 03.12.2016-05.12.2016. • Thirty eligible Fitbit users consented to the submission of personal tracker data, including minute-level output for physical activity, heart rate, and sleep monitoring. • Individual reports can be parsed by export session ID (column A) or timestamp (column B). Variation between output represents use of different types of Fitbit trackers and individual tracking behaviors / preferences. “ FitBit Fitness Tracker Data (CC0: Public Domain, dataset made available through Mobius): License: CC0 1.0 Universal (CC0 1.0) Public Domain Dedication : https://creativecommons.org/publicdomain/zero/1.0/
- 5. Which Fitbit data fits Bellabeat? Bellabeat Features Available Fitbit data Calculates Wellness score ✘ Monitors activity ✓ Monitors sleep ✓ Tracks menstrual cycle ✘ Tracks meditation ✘ Tracks Hydration ✘ Insights into your health & lifestyle ✘ Guidance on how to improve health & lifestyle ✘ Lightweight, safe for skin, and hypoallergenic ✘
- 6. Cleaning up the data and my notes: • The data was cleaned of duplicates, nulls and other common errors. • Fields where transformed to their correct formatting, including dates. What I Noted During Cleaning: • Number of entries is vastly different amongst participants • There is no information about specific models used. (to compare functions, battery life…) • No information about age or gender (Bellabeat is targeted towards women) • Not enough metadata description for the attributes. • Example: what is “sedentary active”? • Does “sedentary minutes” include sleep minutes? Do they overlap? • Dates: 12/04/2016 - 12/05/2016 (dd/mm/yyyy): There is discontinuity in entries for some IDs
- 7. Notes about Activity Data: • Count of unique IDs= 33 -> There are 3 extra participants? was there a mistake in entry? • Check ID length: all are 10 chr long • Min date: 4/12/2016 • Max date: 5/12/2016 • number of days in study description= 30 -> some participants had 31 entries? • No IDs had more than 1 entry/day • There were some entries with zero total steps and zero activity
- 8. Notes about Sleep Data: • Number of Unique Ids: 24 --> There are 6 participants who have no records. • Filter to check length of IDs: all IDs are 10 characters long • Dates: 12/04/2016 - 12/05/2016 (dd/mm/yyyy) --> There is discontinuity in entries for some IDs • Does sedentary time include sleep time? How accurate is the device in differentiating these two? • Some IDs had the 1440 minutes divided between sedentary minutes, active minutes and sleep minutes. • Some had 1440 sedentary minutes + >400 sleep minutes! See next slide
- 9. From Rstudio - https://rstudio.cloud/content/4801669 a) Total time of usage is >1440 minutes: could be due to overlap of sedentary and sleep minutes (recording minutes as both sedentary and sleep)? As a result: I couldn’t use total time of usage in analysis b) Total time of usage is = 1440 minutes:
- 10. • 3 IDs had more than 2 entries/day (not duplicates) this resulted in some IDs having 31 entries • 2 IDs had • In one day: an entry with 2 in TotalSleepRecords+ another entry with 1 in TotalSleepRecords. (Total minutes asleep are greater than allowed by sum of active and sedentary minutes)
- 11. Fitbit: Questions asked 1) What did participants like to use it for more? Track activity, or track sleep? 2) Does usage differ with different days of the week? Working days vs weekends? 3) Is there any difference in average sleep and activity between consistent and non-consistent users? Is there a difference in a specific type of activity? 4) Is there any relationship between sleep and activity that we can use in our marketing strategy? • Eg. By keeping track of your activity and making sure that you’ve had your dose of highly active minutes, you can rest assured that you’ll have some good rest at night! (and you can track that with the device itself!) 5) How long did the participants wear it for? Is it tolerated well? Could this tell us something about the comfort level? Or aesthetics? • This couldn’t be analyzed due to error in total time of use. • We’ll assume that the most consistent users are the ones who like the device and are interested in using it – target group.
- 12. Did they like to track activity? Or sleep? Activity Sleep + 3 - 6 A) Count of Unique IDs Dataset Metadata: number of participants is 30, but.. 33 IDs 24 IDs 30 30
- 13. Did they like to track activity? Or sleep? Activity Sleep B) What did participants track more consistently: Activity or Sleep? Used FitBit every day to track activity Used FitBit every day to track sleep Created on Tableau
- 14. Activity Does Usage differ with different days of the week? Lowest Highest Highest Lowest Sleep Created on Tableau Participants were 41.9% more likely to track sleep on a weekend than on a weekday. Participants were 11.9% more likely to track activity on a weekday than on a weekend.
- 15. Is there any difference in average sleep and activity between consistent and non-consistent users? Define groups of participants according to consistency of usage: • Participants with < 22 entries “non-consistent users” • Participants with >= 22 entries “consistent users” VS
- 16. Consistent Vs Non-consistent Users A) Average Total Minutes Asleep Consistent Users Average sleep = 423.4 minutes (7:03 h:m) Non-Consistent Users Average sleep= 331.4 minutes (5:31 h:m) Difference = 92 minutes (1:32 h:m) Consistent users slept 27.8% more than non-consistent users Average for all participants = 419.2 minutes Average Sleep Minutes Created on Tableau
- 17. Average Fairly + Very Active minutes = 35.5 minutes Average Fairly + Very Active minutes = 21.1 minutes Difference = 14.4 minutes Consistent users had 67.3% more fairly and very active minutes than non-consistent users Consistent Vs Non-consistent Users B) Average Fairly Active + Very Active Minutes Average for all participants = 34.7 minutes Consistent Users Non-consistent Users Average Fairly + Very Active Minutes Created on Tableau
- 18. Average very active minutes = 21.6 minutes Average very active minutes= 11 minutes Difference = 11.6 minutes Consistent users had 96.7% more very active minutes than non-consistent users Consistent Vs Non-consistent Users C) Average Very Active Minutes Average for all participants = 21.2 minutes Consistent Users Non-consistent Users Average Very Active Minutes Created on Tableau
- 19. Relationship between The amount of Activity and Amount of Sleep Can we say: “You can predict how well you’ll sleep tonight by tracking your activity during the day” ? ?
- 20. A) Analyzing (fairly active + very active) minutes on individual days: Sum of Fairly and Very Active Minutes in 1 Day Total Minutes Asleep in 1 day When the sum of fairly and very active minutes in one day is between 100- 275, sleep becomes directly proportional to it. Research paper: “The effect of physical activity on sleep quality: a systematic review” https://www.tandfonline.com/doi/fu ll/10.1080/21679169.2019.1623314 1) Fairly + Very Active Minutes Vs Sleep Created using RStudio – ggplot2 package
- 21. • When the average of fairly and very active minutes is between 25-65 minutes, sleep is directly proportional to it. • If it’s between 70-124, it becomes inversely proportional to sleep. Average Fairly and Very Active Minutes per user Average Minutes Asleep per user B) Analyzing the average (fairly active + very active) minutes vs average sleep for each user through the whole period: 1) Fairly + Very Active Minutes Vs Sleep Created using RStudio – ggplot2 package
- 22. A) Analyzing fairly active minutes vs sleep on individual days: When fairly active minutes is between 0- 143, sleep is inversely proportional to it. 2) Fairly Active Minutes Vs Sleep Fairly Active Minutes in 1 Day Total Minutes Asleep in 1 Day Created using RStudio – ggplot2 package
- 23. • When the average of fairly active minutes is between 17- 37 minutes, sleep is directly proportional to it. • If it’s between 37- 62, it becomes inversely proportional to sleep. Average Fairly Active Minutes per user Average Minutes Asleep per user B) Analyzing the average fairly active minutes with average sleep for each user through the whole period: 2) Fairly Active Minutes Vs Sleep Created using RStudio – ggplot2 package
- 24. A) Analyzing very active minutes vs sleep on individual days: Very Active Minutes in 1 Day Total Minutes Asleep in 1 day When very active minutes in one day is between 30 and 210, sleep is directly proportional to very active minutes. 1) Very Active Minutes Vs Sleep Created using RStudio – ggplot2 package
- 25. • When the average of very active minutes is between 19-70 minutes, sleep is directly proportional to very active minutes. • If it’s between 75- 102, it becomes inversely proportional to sleep. Average Very Active Minutes per user Average Minutes Asleep per user B) Analyzing the average very active minutes, with average sleep for each user through the whole period: 2) Very Active Minutes Vs Sleep Created using RStudio – ggplot2 package
- 26. What can we infer? What we’ve learned Possible Questions Action Recommended 1) Device was used more frequently and consistently to track activity rather than sleep. 2) Users tend to track sleep much more on weekends than weekdays. 3) There wasn’t not much difference in activity tracking between weekdays and weekends. • Are participants more interested in tracking activity than in tracking sleep? • One can guess hours of sleep but not count of steps • The cause for less sleep tracking? • battery running out at the end of the day?. • an issue of comfort? • Weekends: Better chance of charging for sleep tracking? • Weekends: Less worry of not getting good sleep due to comfort? 1) Focus on battery life: it will last x days on a single charge of x hours. (no need to worry about charging it in the middle of the day/ before going to sleep) 2) Focus on comfort level of the watch, or that it can be used while not worn? 3) Focus on tracking activity for this product, and focus on sleep when promoting other more suitable products.
- 27. What can we infer? What we’ve learned Possible Questions Action Recommended 4) Consistent users slept 27.8% more than non- consistent users 5) Consistent users had 67.3% more fairly and very active minutes than non-consistent users 6) Consistent users had 96.7% more very active minutes than non- consistent users 7) When the sum of fairly + very active minutes > 100, sleep becomes directly proportional to activity. • Did consistently wearing the device motivate the users to take better care of their sleep and activity? • Or were the participants with pre- existing better sleep and activity more motivated to wear it? • We can say that by tracking your sleep and activity consistently, you’ll be more motivated to take care of your sleep and activity levels. (example: wearing nice & new gym wear to motivate yourself to exercise.) • Mention the scientifically-proven relationship between sleep and activity. “You can predict how well you’ll sleep tonight by tracking your activity during the day”?
- 28. Thank you for taking the time to view my work! Hiba Shawa | November 7th 2022 Got questions? Please contact me on: shawahiba@gmail.com
Editor's Notes
- #27: Note: A fully-charged Charge 5 has a battery life of up to 7 days. Charging fully takes about 1-2 hours.