Analysis and Insights
Visualizations, Exploratory Analysis, and Statistical Analysis
a high level overview
The first step of my analysis was to see the initial reaction online on the launch dates of Fenty Beauty Pro Filt'r foundation and Tarte Shape Tape foundation. I used Google Trends and the respective plots are below.
Rihanna launched Fenty Beauty on September 8th 2017, and as shown above, there is a significant uptick in the number of hits for the word "Fenty"on that day.
On January 16th 2018, Tarte Cosmetics released pictures of swatches of the highly anticipated Shape Tape Foundation. Tarte Cosmetics was wrapped up in the social media storm in the following month as YouTube influencers and Twitter users continued to react to the product's limited shade range.
ANALYSIS OF YOUTUBE COMMENTS
I scraped the comments under the YouTube videos of the following beauty influencers:
I created a word cloud of the most recurring words in the YouTube comment section. Above are the words under two YouTube videos on Fenty Beauty. We see positive words like "beautiful," "love," "skin," and even "Rihanna." Most of the words highlight the generally positive reception to Fenty Beauty.
On the other hand, for Tarte Cosmetics, we see a lot of people complaining about the lack of shade range. The word "shade(s)", "range", "inclusive," and "dont" occur very frequently. What is also interesting is that "fenty" also comes up as a popular word, as lots of people made comparisons to the extensive shade range of Fenty Beauty. The comments aren't all negative as we see positive words like "amazing," "speaking," and "platform," as YouTube users praise the beauty influencers for using their platform to call out a reputable brand.
I also conducted a sentiment analysis of the YouTube comments. The graph above shows that the majority of comments were positive at about 34%, followed by comments expressing joy at 24%. Anticipation hovers at around 12%, and negative comments make up about 6%.
For Tarte Cosmetics, positive comments make up about 27%. Again, I hypothesize that a considerable portion of positive comments come from people praising the YouTube influencers for speaking out. Negative comments are notably higher, at 11%. Comments expressing joy are significantly lower, at about 14%. We also see anger, disgust, and sadness climb to 6%, 5%, and 7% respectively.
Above is a numerical sentiment analysis of Fenty Beauty comments using the "bing" method. We see a stronger positive skew for Fenty Beauty when compared with Tarte Cosmetics below.
exploratory Analysis of sephora Foundation reviews
For the first step of this part of my analysis, I wanted to get a picture of the overall foundation reviews on Sehpora segmented by skin tone. My initial hypothesis was that darker skin tones would be underrepresented in reviews, but I did not expect it to be this significant.
The vast majority of reviews come from those with fair, light, and medium skin tones. Ebony is almost non existent, and the other shades hover under 5%. One can reasonably infer that the lack of shade diversity in many make-up brands largely explain these findings. This proves that while there have been major steps towards diversity and inclusion in the beauty industry, there is still so much room for improvement.
The 40 shade range Pro Filt'r foundation appears to be well received, because for Fenty Beauty, we see a notable increase representation from darker skin tones. One thing that particularly stands out to me is the significant increase in representation of porcelain skin tones. This shows that developing foundations that reflect the full range of customers is not only about catering to people of color, but also, people with the lightest skin tones benefit as well.
Before I conducted my statistical analysis, I wanted to see if a visualization could show a relationship between skin tone and the likeliness to recommend foundation. Based on the chart above, there seems to be a relationship. People with darker skin tones appear to be more willing to recommend beauty products.
STATISTICAL Analysis of Fenty beauty Foundation reviews
Is It Possible to Predict Whether a Person Will Recommend Foundation Based on their Skin Tone?
I first converted skin tone to a new variable of factors using the as.factor() function. The skin tone omitted is "dark," which happens to be the second darkest skin tone. Thus, the linear model above shows the effects of skin tone on willingness to recommend, relative to the omitted "dark" skin tone. We see that all the skin tones but "deep" and "ebony" have statistically significant effects (p < 0.05) on recommendation relative to "dark." This makes sense because "dark," "ebony," and "deep" are the three darkest shades. Porcelain, the lightest shade, has the largest negative coefficient. Unsurprisingly, "fair" and "light "have the next largest negative coefficients.
​
While the model is statistically significant, the R-squared is on the low end, indicating further room for improvement.
We also see the same results when we make rating the explanatory variable. This model appears to be stronger because of the larger R-squared. Also the estimates of the skin tone factors are also significantly greater than the previous model.
Building a logistic model for predictions...
I initially planned to use skin tone as the only independent variable, however I decided to develop another feature to improve the predictability. I found the words most highly correlated with skin tones (I assigned the numbers 1-8 to the skin tone categories, 1 being "porcelain" and 8 being "ebony"). I named this feature "toned words." The words are shown in the word cloud above.
The logistic regression model correctly predicts customer recommendations about 78% of the time. It appears that we can predict a customer's willingness to recommend Fenty Beauty products based on features related to skin tone.
CONCLUDING ISIGHTS AND takeaways
-
It is clearly beneficial for companies to develop product collections that reflect the full spectrum of consumers.
-
Customers are generally more receptive to brands that have diverse and correct ranges of shades and skin tones.
-
There appears to be a statically significant relationship between skin tone and willingness to recommend products with diverse shade ranges.
-
Rihanna’s collection not only provides an impressive shade selection, it also highlights the need for other reputable cosmetic brands to take up the mantle when it comes to diversity and inclusion.
​
***I was unable to scrape Tarte Cosmetics Shape Tape reviews because the product has been discontinued and pulled from all online platforms.
The first independent binary variable is skintonebinary, this takes the value 1 if the skin tone is olive, deep, dark, or ebony. The second independent binary variable takes the value 1 if there are any "toned words" in the text of the reviews.