title: "Bellabeat" output: word_document: default html_document: default date: "2023-07-12"

knitr::opts_chunk$set(echo = TRUE)

Case Study: Bellabeat

Bellabeat is a manufacturer of health focused products for women. In order to improve its growth potential, Bellabeat would like to gather data on how non bella beat users use their smart devices to get better insight on the market as a whole.

Business Task:

Gather data on how non bella beat users use their smart devices. Figure out how to apply the data and trends gathered to Bellabeat users. Create a marketing strategy based around the data.

Prepare:

Data is stored on a local hard drive and is organized in a combination of the wide and long format. The data sources used in this analysis are from an Amazon fitbit survey distributed from 3.12.2016 - 05.12.2016. The dataset contains personal tracker data, including: minute-level output for physical activity heart rate Calories burned Distance traveled Intensity of output sleep monitoring METs

Process:

Each document was imported into sheets. They were then checked for duplicate entries and three were found in the sleep dataset and deleted. I then checked for blank entries and none were found. I noticed that there were a lot of decimal places in most files. Each decimal related column was reduced to 1-2 decimal places. The METs had an additional 0 in each entry so I divided one cell by 10 and applied that formula for the rest. Dates and times were separated into columns. The times were formatted to plain text so that they could be concatenated with the AM and PM column. Once combined, the time column was reformatted back to the time format. Each column was trimmed to delete any extra spaces in the front or back entries. ID numbers were checked for consistent length and unnecessary characters. The data has been properly formatted to date and time format with the date and time separated for easier reading of the data. I also converted the sleep minutes into hours and calculated the minutes it took to fall asleep by subtracting the minutes in bed from the minutes asleep.

Analyze/Share:

There are 33 participants in this survey, but a few datasets did not have enough data to extract any significant insight on the group. The weight dataset had 8 unique entries and the heart rate dataset had 14. Since these datasets represent less than half of the group as a whole I will not be using them for most of my analysis. The datasets I will be focusing on for this analysis are: Sleep Daily Activity Distance Calories METs

The CDC cites that the average American walks 3,000 to 4,000 steps a day or 1.5 to 2 miles a day. The average steps taken by the fitbit users in the study was 7,638 steps. If 2,000 steps equals one mile, the average user walked 3.8 miles daily. That is considered active but not ideal. The common benchmark to strive for is 10,000 steps. Out of the 33 participants only 7 clocked an average of 10,000 steps or higher during the 30 days the study took place. That is only 21% of the participants. According to the National Institute of Health, taking less than 4,000 steps is considered light activity. Six participants recorded less than 4,000 steps, with two dropping below 2,000.

The average sedentary minutes was pretty high at 991 minutes or 16 hours. This is pretty high considering the national average sedentary time is 9.5 hours. The average light activity minutes was 192 minutes which equals a little over 3 hours. Light Activity is defined as any activity between 1.5 - 3 METs. This is the equivalent of a light walk or standing in line at the grocery store. The average fairly active minutes was 13.5 minutes and the average very active minutes was 22 minutes. These two categories are considered to be “active” by fitbits standards because they fall in the 3 or greater MET range. For context, a 4 is a brisk walk, a 6 is a light jog, and a 9 is an intense run. When we combine the average fairly active minutes and very active minutes we get an average “active” time of 35.5 minutes.

Next, I converted the dates into the days of the week to see what days had the most active minutes. When it came to light activity minutes most people were lightly active on Friday and Saturday, both averaging 204 and 207 minutes and the closest next day being Tuesday at 197 and Monday at 192. This makes sense because light activity has been defined as standing or lightly walking. Friday and Saturday are the days people often attend events or go out for entertainment where they either stand or walk around.

I combined the fairly and very active minutes to get the “Active” time as defined by Fitbit. I found that the days with the highest average active minutes were Monday, Tuesday, and Saturday each averaging 37 minutes of activity above 3 METs. The lowest active days were Thursday and Friday. A consistent trend is that the partici[ant’s activity levels are at their highest Monday and Tuesday but begin to lower Wednesday through Friday but see a spike on Saturday and drop a little Sunday. I then filtered out METs less than 3 and found that the average MET was 5.3. This means the average MET level reached when a person is “active” is 5.3 METs.

Another interesting finding was that the average time to fall asleep was 39 minutes. According to the sleep foundation, the average person falls asleep within 15 to 20 minutes of lying in bed. Bellabeat users on average are taking about 20 minutes longer than the average person to fall asleep. Of the 24 participants with sleep data, 16 averaged more than 20 minutes to fall asleep with the highest average being 309 minutes (5 hours). One reason why the users may be experiencing longer wake periods on their way to sleep is because of the high sedentary time mentioned earlier. “Sedentary people do not get the boost in sleep quality that comes from regular activity” as stated in a Washington Post article explaining the increase in the amount of time the average american sits per day. A trend that I observed was that the minutes that it took to fall asleep decreased as the week went on, peaking Sunday at 51 minutes, and decreasing steadily until Thursday where it reached a low of 33 minutes, then slightly rising on Friday and Saturday to 40 minutes.

Thursday is the day that takes the least amount of time to fall asleep and is also the day that has the least active minutes. Though these two metrics share the same day of the week for their lowest metrics, it does not necessarily mean there is a correlation between sleep and activity for the fitbit users. The average user takes the longest to fall asleep on Sunday but that isnt the most active day on average. On average the most active days are Monday, Tuesday and Saturday.

Act:

Alert users to take more steps before they end the day Users average 7638 steps a day, that's not too far from the recommended 10,000 steps suggested by the CDC. If they got an alert of how many steps they’ve taken in the day and were encouraged to at least try to get that number to the nearest thousand that might strive to get to 10,000 steps. Users could be alerted if they have taken less than their average steps for the week and can then be encouraged to take the necessary steps to beat that average or maintain it. Being alerted that they are regressing may be enough motivation to encourage users to get active and stay consistent. Sleep To help users sleep better, Bellabeat could encourage users to engage in a relaxing activity like yoga before they go to bed. This will help tire and/or relax the body for bed to allow for an easier passage to dreamland.

DevGitsit / Bella

BellaBeat Report #3