ULTA's own labels or other brands? Web Scraping Project

Yunmei Zhang
Posted on Feb 2, 2020

Background 

Founded in 1990, Ulta Beauty has gained the favors of many beauty shoppers and become one of the largest U.S. beauty retailers for cosmetics, skin care products, hair care products, fragrance and salon services. As of November of 2019, Ulta Beauty operates 1,241 retail stores across 50 states and also distributes its products through its website. It carries both in stores and online with more than 25,000 products from approximately 500 beauty brands across all categories and price ranges, including Ulta Beauty’s own private label. 

 

The Goal 

As a beauty shopper myself, I particularly enjoy the convenience and variety that Ulta's beauty offers. I no longer need to spend tons of time in beauty counters located inside of a mall, blindly choose the products from a limited selection of brands, most of which are luxury brands that are beyond my budget. Thanks to Ulta and many other beauty retailers, I am now exposed to both well-established brands and some emerging beauty brands with a much wider price range. Most importantly, I can read the reviews of any products that I am interested in before I take the risk of trying them on. Nowadays, I find myself hardly buy any products without reading the reviews ahead and figuring it out whether it is a good fit or not. So I decide to scrape Ulta's products' ratings to find out whether customers prefer Ulta's own labels or other well-established brands. 

 

Approach

I chose to extract skincare and makeup products data from Ulta's website using scrapy package in python. I built a scrapy spider for each top category (skincare and makeup). Below graph is a tree diagram of my skincare scrapy spider. Makeup scrapy spider is very similar to this, except for it has fewer product categories, 8 categories in total. The information I extracted includes the product category, brand, name, average rating, total reviews, price, and description. 

Data Cleaning 

I used python to clean the data I scraped from Ulta. The main challenge I encountered was missing values due to no ratings/reviews for some products. I dropped the products that have missing ratings/reviews, and also filtered out the products that have less than 10 reviews. This reduced my sample size from 9417 to 6310. Though 10 ratings is not a good representation of general's view of the product and could be biased due to some extreme reviews, due to the limited reviews on Ulta's website, I decided to use 10 reviews as a cutoff so I have enough products to compare them with. 

 

Analyses

Two-sample t-test On Skincare And Makeup Products' Average Ratings

The first analysis I did was to check if there is a significant difference between the means of skincare product rating and makeup product rating using a two-sample t-test in python scipy package. A two-sample t-test is a statistical testing performed on two random samples that are independently obtained from two populations. The purpose of the test is to determine whether the difference between these two populations is statistically significant.

My null hypothesis(H0) was that they had the same means, and my alternative hypothesis (Ha) was that their means were different. The p-value I calculated, which is the probability of obtaining the observed results of a test, assuming that the null hypothesis is correct, was much smaller than the significant level of 0.05. This suggests that there is less than 5% chance that the null hypothesis is right. Therefore, I rejected the null hypothesis and concluded that there is a significant difference between the mean of skincare product rating and the makeup product rating. Based on this result, my following analyses would treat these two categories separately. 

Distributions Of Ratings Among Product Sub Categories

I then plotted the distribution of ratings among different product sub categories for both skincare and makeup products using box plot. A box plot is a standardized way of displaying the data set based on a five-number summary: minimum, first quartile, median, third quartile, and maximum. A box is from the first quartile to the third quartile. A vertical line goes through the box at the median. The whiskers go from each quartile to the minimum or maximum.

Chart on the left is a box plot for the skincare products with 13 categories in total, and on the right is a box plot for the makeup products with 8 total categories. Both plots are sorted by the median in the descending order. We can see from the plots that in both skincare and makeup categories, ULTA Collection has the lowest median rating among all sub categories. 

Comparison On Mean Rating And Price Among Product Sub Categories

Does price has an impact on the rating? How does Ulta's own brand stand in price compared to other brands? I then computed the average rating and average price of each sub category and plotted them in the following graphs for comparison. 

The left graph in turquoise color is for the skincare product categories and the right graph in salmon color is for the makeup product categories. For both graphs, average rating is plotted using bars and sorted by descending order. The scale for the product rating is shown on the left y-axis. Average price is plotted using blue dots and the scale is shown on the right y-axis. 

At a glance of the charts, average rating does not seem to have a strong correlation with the average price. Supplements from Skincare product has the highest average rating even though its average price is the second highest among all skincare sub categories. It appears that for makeup products, there is a slightly stronger correlation between the two. Then I used Pearson method to calculate the correlation between average rating and average price for skincare product and makeup product to validate my observations. The coefficients were 0.092 and 0.17 in skincare and makeup categories, respectively, indicating a weak correlation between price and rating. 

What is interesting to me is the fact that not only does ULTA Collection has the lowest average rating (3.95 for skincare, 4.03 for makeup), it also has the lowest average price among all categories for both skincare and makeup products. A question came to my mind: what is the distribution of rating like for ULTA Collection products? Whether this low average rating is due to a few poorly rated products or customers are consistently leaving poor reviews among all ULTA Collection products? So I plotted two histograms to show the distribution of all ULTA Collection products' rating for both skincare and makeup. 

Distribution Of ULTA Collection Products' Rating 

Both histograms show a wide spread of rating among ULTA Collection products. Both graphs are skewed to the left, suggesting that indeed there are some poorly rated products that could hurt the average rating of all ULTA Collection product. On the other hand, there are a few products that are highly rated. In order to improve ULTA Collection overall's rating, it will be important to find out the cause of this variability. Is Ulta delivering the same quality of product to its customers? What product line is Ulta strong at and how can it improve the other product lines?

Conclusion

In summary of my analyses, customers prefer other brands more than Ulta's own labels in both skincare and makeup products. Price does not play a big role in product's rating, Ulta should not primarily focus on pricing strategies in order to gain customers' favors. Instead, it should put more focus on market research and collecting  customers' voice to improve its products that will best suit their needs. An easy start is to encourage more customers to leave reviews on its website as I found the amount of reviews were limited when I collected the data. Ulta should continue expanding partnerships with more brands and products while developing its own products. 

For future works, I would be interested in collecting more information on reviews to find out what customers like or dislike about a product. Product ingredients is another useful information to identify any trend in clean/healthy ingredients vs non-clean ingredients. 

 

Thank you!

Thank you for taking the time to read my project. Here is a link to my GitHub repository if you would like to explore my code. Please feel free to reach out to me if you have any questions or want to discuss more. 

 

About Author

Yunmei Zhang

Yunmei Zhang

Yunmei (May) graduated from Cornell University with a master degree in chemical engineering. After graduation, she started a career as a process engineer in a manufacturing company, where she utilized the power of data in problem solving and...
View all posts by Yunmei Zhang >

Leave a Comment

No comments found.

View Posts by Categories


Our Recent Popular Posts


View Posts by Tags

#python #trainwithnycdsa 2019 airbnb Alex Baransky alumni Alumni Interview Alumni Reviews Alumni Spotlight alumni story Alumnus API Application artist aws beautiful soup Best Bootcamp Best Data Science 2019 Best Data Science Bootcamp Best Data Science Bootcamp 2020 Best Ranked Big Data Book Launch Book-Signing bootcamp Bootcamp Alumni Bootcamp Prep Bundles California Cancer Research capstone Career Career Day citibike clustering Coding Course Demo Course Report D3.js data Data Analyst data science Data Science Academy Data Science Bootcamp Data science jobs Data Science Reviews Data Scientist Data Scientist Jobs data visualization Deep Learning Demo Day Discount dplyr employer networking feature engineering Finance Financial Data Science Flask gbm Get Hired ggplot2 googleVis Hadoop higgs boson Hiring hiring partner events Hiring Partners Industry Experts Instructor Blog Instructor Interview Job Job Placement Jobs Jon Krohn JP Morgan Chase Kaggle Kickstarter lasso regression Lead Data Scienctist Lead Data Scientist leaflet linear regression Logistic Regression machine learning Maps matplotlib Medical Research Meet the team meetup Networking neural network Neural networks New Courses nlp NYC NYC Data Science nyc data science academy NYC Open Data NYCDSA NYCDSA Alumni Online Online Bootcamp Online Training Open Data painter pandas Part-time Portfolio Development prediction Prework Programming PwC python python machine learning python scrapy python web scraping python webscraping Python Workshop R R language R Programming R Shiny r studio R Visualization R Workshop R-bloggers random forest Ranking recommendation recommendation system regression Remote remote data science bootcamp Scrapy scrapy visualization seaborn Selenium sentiment analysis Shiny Shiny Dashboard Spark Special Special Summer Sports statistics streaming Student Interview Student Showcase SVM Switchup Tableau team TensorFlow Testimonial tf-idf Top Data Science Bootcamp twitter visualization web scraping Weekend Course What to expect word cloud word2vec XGBoost yelp