Scraping and Analyzing ELC Brand Presence on Sephora
We all celebrated Mother’s Day on May 13th, thanking the beautiful mums in our own ways. I was strolling around in NYC, enjoying a beautiful summer day and noticed many beauty brands attracting consumers through promotions and gift packages for moms. It made the data enthusiast in me want to go behind the scene and look at beauty in terms of numbers. I decided to limit my scope to a beauty giant based here out of NYC – The Estee Lauder Company(ELC). And where else to look for data than every girl’s favorite beauty specialist – Sephora. ELC owns 29 beauty brands, of which 19 are available on Sephora. I decided to look into these 19 brands.
Data Scraping and Cleaning
The first and foremost step was to scrape Sephora website to collect data. The website lists down cost, number of reviews, customer ratings and number of likes (favorites) for each product. The website has products divided into various categories (Makeup, Skincare, Fragrance, Hair, Bath & Body and so on) for easy search by consumers. Scraping the website also gives you total number of products for each brand by category.
However, what the website does not give you is sales figures (for obvious reasons) for brands. The data scientist in me craved more data than what I could get from Sephora to have at least some indication of sales for ELC. Thanks to Beauty Packaging, I could get insights into rankings and sales of the beauty companies. I scraped the website to understand ELC’s position in the industry. As there are multiple beauty categories, I wanted to get some sense on which ones are more profitable for ELC and what is driving sales. I could find this data in ELC’s 2017 annual report.
I used Selenium to scrape both Sephora and Beauty Packaging to extract the required data. One particular thing to be taken care of while scraping Sephora was that “Ratings & Reviews” section on the product page shows up only when scrolled down to the place on page where it is hidden. This section has the totals no. of reviews and rating for each product, which I needed to scrape. I made sure my Selenium code scrolled down and got the Ratings & Review section showing up on each product’s page.
Since I just needed 11 rows of data from ELC annual report, I just grabbed them manually. Once I had all the required data, I decided to use Python for data cleaning, processing and analyzing.
Data Analysis & Insights
I started off by looking at the sales of companies. ELC ranked 3rd biggest beauty empire globally in 2017 with sales of USD 11.8 billion. I looked at the trends of top 6 beauty companies (top 6 based on 2017 sales) for past 7 years.
A little research into these trends: P&G has been divesting many of its brands to focus on selective categories. They sold more than 40 brands in 2016 to Coty and hence the increase there for Coty. ELC’s acquisition strategy has been helping it to grow and surpass P&G marginally in 2017 to become number three. Also, compared to other top beauty organizations, ELC has a unique market position that it only targets the high-end/ luxury market segment vs. drugstore segment.
Looking at ELC’s annual figures for 2017 by categories, makeup was the biggest driver of sales, followed by skincare. In terms of operating margins, skincare had the highest margins.
Deep diving into data scraped from Sephora across all ELC brands, Clinique and Tom Ford have the strongest brand presence. I defined presence as the number of products available to consumers by brand.
Next, I looked at the cost range for each brand to get an insight into the costlier vs. the more affordable brands by ELC. This was as expected: brands like Glamglow, Clinique, Too-Faced have comparatively lower median product costs, while niche brands like Kilian and LaMer have high median costs. Kilian, the prestige fragrance brand, is the costliest of all the ELC brands on Sephora with median cost of USD 235 across products. Glamglow has the lowest median price of USD 24.
Does cost influence customer engagement?
I also wanted to look for an answer to another question: all ELC brands already operate in luxury segment but are the customers, who are ready to shell out more money on luxury brands vs the drugstore, still sensitive to cost? For this, I looked at customer engagement against the product costs by brand and by each category. Customer engagement was defined as the average number of reviews across products in a category for a brand. This gives an idea into how many customers experienced the brand. Number of reviews don’t equate to exact no. of customers who have used the product but given the limited data available, no. of reviews provides some indication of the user base. Product costs are averaged across products for a brand in a category. I plotted scatterplots for customer engagement (log) vs. cost. Here are the findings:
- Skincare – Beyond certain price point ($40), generally customer engagement decreases with price
- Makeup – Generally beyond $30 price point, customer engagement decreases with product cost
- Fragrance and Men – again, generally with increase in cost, engagement seems to fall down
Is there any impact of brand presence and cost on avg. rating of a brand?
This was another obvious question I wanted to answer. I plotted bubble charts, where bubble size represents average cost of the brand. Below are the findings by category:
- Skincare, Fragrance and Men – Cost and presence does not seem to influence ratings
- Makeup – Higher cost brands seem to have higher ratings. Brand presence apparently does not differentiate the ratings
Likes vs. Ratings: is there a different story?
Another interesting thing that I explored was if average no. of likes for a brand reflect the ratings. Sephora website allows customers to add products to their ‘loves’ list. Any product a customer loves/likes can be added to this list and accessed later on. This can be considered similar to ‘likes’ on Facebook. Surprisingly, brands with higher no. of likes have comparatively lower ratings across categories in general.
- ELC positions itself different from other beauty companies by targeting luxury segment only
- Makeup is the biggest contributor to driving sales for ELC, while skincare has the highest operating margins
- Clinique and Tom Ford have the biggest brand presence at Sephora
- Kilian and La Mer are the costliest (median and highest cost) of all the ELC brands
- Cost seems to influence customer engagement, reflecting on customers being cost conscious
- Generally, there seems to be an inverse relationship between likes and avg. ratings
- Other types of data such as sales at product level, customer profiling, Sephora footprint and reviews analysis can be used to further analyze ELC brands and make cross-company comparisons