Analyzing Product Ratings and Reviews for Various Furniture Companies

Jason Liebmann
Analytics Vidhya
Published in
9 min readMay 1, 2020

--

Photo Courtesy of Apartments.com: https://images1.apartments.com/i2/KOBkf5EUHWgcJ6oNIvyT-BFIA0_fDZiunX6u8M_Vnkk/111/stonegate-luxury-furnished-apartments-mesa-az-1-bedroom.jpg

Imagine you just bought your dream house, but it is completely empty. It may seem exciting to have a completely empty canvas to work with and to be able to make your new house into whatever you can imagine; however, the process can quickly become daunting. A simple google search for Ikea, Wayfair or Ashley’s Furniture can easily reveal just how many options there are and the process of furnishing your new place can become overwhelming in a short period of time.

On top of all the options, it can be difficult to know which products are truly the best without trying them out and testing them in person. However, going to visit stores in person can make the entire process even more tiresome than beforehand. Online product reviews can help mitigate this issue since they give other customers a perspective on different aspects of the product from people who supposedly bought them. However, it can sometimes be difficult to know which reviews to trust as well as compare ratings and reviews across platforms.

This post analyzes customer product review data from Ikea, Wayfair, Ashley’s Furniture and Haverty’s in order to help you more easily narrow down your furniture search as well as get a better grasp of what these reviews truly mean.

1. DISTRIBUTION OF PRODUCT RATINGS ACROSS WEBSITES

Customer reviews are becoming more and more popular each day as customers demand more transparency on the part of companies and companies attempt to deliver that transparency in order to stay competitive. As more and more companies offer review insights into the products, it becomes harder and harder to distinguish between them. Most importantly, it can be very difficult to know what exactly a given star rating indicates about the product. For example, a 4 out of 5 may mean one thing on one website but may mean something completely different on another. Additionally, it can be hard to know if a 4 out of 5 is good or not for a given website as average ratings can vary greatly across different sites.

Boxplots of Product Star Ratings by Company (the blue diamond represents the mean product rating for that company)

Looking at Ashley’s Furniture, Haverty’s, Ikea, and Wayfair, we can see that their product ratings certainly vary quite a lot. From the graph, it is clear that Haverty’s has the highest average product rating, followed by Ashley’s, Wayfair and finally Ikea. These differences could suggest two things: on average some of these companies simply offer better products than others or there are systematic differences in the reviews on these sites. Looking at the distribution of product ratings for each company may give us some insight.

Both Haverty’s and Ashley’s have very few poor product ratings on their site, while both Ikea and Wayfair suffer from quite a few. It is possible that Ikea and Wayfair have some bad products; however, given the success of the two companies, it appears more likely that Ikea and Wayfair customers are more likely to give poor reviews when products deserve it, making their good reviews mean more. On the other hand, Ashley’s and Haverty’s customers seem to leave very few poor reviews of products making their good reviews hold less weight since Ashley’s and Haverty’s almost certainly have some poor quality products for sale.

2. FREQUENCIES OF 5-STAR RATINGS & “GOOD” PRODUCTS

Taking a deeper look at the product reviews, specifically 5-star reviews, one can notice a lot of variation in the percentage of products that have 5-star ratings on the various websites. When examining all reviews (the graph on the left), we can see that Haverty’s has the highest proportion of 5-star ratings. However, as we can see when looking at product ratings with 5 or more reviews (the graph on the right), only 12% of Haverty’s products have a 5-star rating. Therefore, we can better understand that many of Haverty’s 5-star ratings are due to product reviews by a small number of people and therefore may not be very reliable.

Bar Charts of Proportion of Ratings that are 5-Stars for All Ratings (left) and only Ratings with 5 or More Reviews (right) by Company

We can see that product reviews on Ikea and Wayfair’s websites follow a similar trend to Haverty’s suggesting some of their 5-star ratings are due to a few avid fans rather than mass approval as well. However, Ashley’s Furniture bucks the trend of the other three and actually sees an increase in the proportion of 5-star ratings on their site when only considering products with 5 or more reviews. This suggests that many of Ashley’s 5-star product ratings are well agreed upon and are more reliable than any of the other companies’ 5-star ratings. Furthermore, this implies that many of the bad product ratings on Ashley’s website are from small numbers of people suggesting those ratings may simply be based upon a few bad experiences and may not be very reliable.

Furthering this analysis, we can also investigate if similar trends hold true for “good” products (which are considered to be products with a rating above 4.5 stars). The graph below provides this analysis.

Bar Charts of Proportion of Ratings that are above 4.5-Stars for All Ratings (left) and only Ratings with 5 or More Reviews (right) by Company

As we can see, Haverty’s once again has the highest proportion of “good” products on their site; however, they suffer the greatest drop in proportion of “good” products when examining only those products with more than 5 reviews. This tells us that “good” Haverty’s product ratings are most likely unreliable and are not very trustworthy. Another interesting takeaway to note is that while Ikea and Wayfair had major dropoffs in their 5-star ratings when looking only at products with 5 or more reviews, their “good” product ratings overall do not follow that same trend as their proportions only drop slightly or stay the same. This suggests that their “good” product reviews that are not 5-star ratings are as a whole more trustworthy and more common among their customers than their 5-star ratings. Finally, we can see that Ashley’s proportion of “good” product ratings again increases when looking only at products with 5 or more reviews suggesting these high ratings are well shared amongst many of their customers.

3. ARE ALL PRODUCT RATINGS EQUALLY TRUSTWORTHY/SHARED AMONGST CUSTOMER BASES?

Plain and simple, no. There is a massive amount of variation in terms of the number of people that leave reviews on these companies websites, both in terms of the average number of reviews per product as well as the distribution of the number of people that leave reviews for different products for different companies.

Bar Chart of Average Number of Reviews per Product (left) and Boxplot of Number of Reviews for Each Product (right)

From the graphs above, we can see that Ashley’s, Haverty’s, and Ikea all have a similar number of reviews per product on their websites as they vary by at most 5 people on average. However, looking closer at their distributions on the right, we can see that Ikea has more products with a higher number of reviews than Ashley’s or Haverty’s since Ikea’s distribution of the number of reviews per product stretches higher up than the other two distributions.

The main discrepancy we notice in the graph above; however, is Wayfair versus all of the others. Wayfair has on average more than 10x as many people review their products than any of the other furniture stores in this analysis. We can also see Wayfair’s distribution on the right stretches way higher than any of the other three distributions, showing that Wayfair has many products with a ton of reviews. This may be in part due to the fact that Wayfair is all online, making it more convenient for people to leave reviews online. However, this is also most likely in part to Wayfair’s customers being more vocal about their company’s products.

Regardless of the reason, the fact that Wayfair has so many more reviews on their website makes their product ratings much more trustworthy than any of the other companies’ product ratings. If a company were to pay people to write reviews, it would be much more costly to get almost 300 reviews per product on your website than 25 reviews per product. Therefore, the sheer amount of reviews on Wayfair’s website makes their product ratings more trustworthy and believable.

4. DOES PRICE INFLUENCE RATING?

Whether surprising or not, depending on if you believe people incorporate the price of a product into their review of it, the short answer is yes. The more complicated answer is yes, and in different ways depending on the company. From the graph below, we can see that a product’s price can have vastly different effects on their rating depending on the company they are being sold from.

Scatterplot of Price of Product vs. Rating of Product by Company with a Linear Regression Overlayed

For Ashley’s Furniture and Haverty’s, the price has little effect on the rating, but if anything, on average, the higher the price of the product, the lower its rating. On the other hand, price has a bigger effect on product rating for Ikea and Wayfair, albeit in opposite directions. For Ikea, the more expensive products see a drastic decrease in their rating; however, this is mostly driven by the fact that Ikea has a lot of 1-star reviews for a good number of their more expensive products. Wayfair is the only company of these four companies for which their product ratings seemingly improve the more expensive the products get. Therefore, it appears that Wayfair is the only company of these four at which you can spend more and not fear getting ripped-off at as much as if you were to shop at any of the other companies.

5. DOES THE NUMBER OF PRODUCT REVIEWS DETERMINE ITS RATING?

Once again, the short answer is yes and the more complicated answer is yes but in different ways for different companies.

Scatterplot of Number of Reviews vs. Rating of Product by Company with a Linear Regression Overlayed

The graph above clearly shows that the number of reviews for a product impacts that given product’s rating in very different ways for the four companies under analysis in this study. For both Ashley’s and Haverty’s, the more reviews a given product has the lower the product’s rating on average. However, due to the fact that neither Ashley’s nor Haverty’s has that many reviews on any of its products, the strength of this negative relationship is very weak for extrapolation for larger numbers of reviews as shown by the large confidence intervals for the slope for these two companies when you get to large numbers of reviews.

For Ikea, the relationship between the number of reviews and the product ratings seems to be positive. This relationship is also not the strongest for extrapolation to larger numbers of reviews either as shown by its large confidence interval for large numbers of reviews; however, it is stronger than either Ashley’s or Haverty’s relationships.

Finally, Wayfair has by far the strongest relationship for extrapolation to large numbers of reviews since it has a large number of reviews on its products. While Wayfair’s relationship is by far the most defined in the data due to its much larger distribution of the number of reviews for its products relative to the other companies, Wayfair’s product’s ratings do not seem to be impacted at all by the number of reviews the product has.

In order to determine what to shop for on each website, we will look at customer reviews in aggregate for each product category and will then compare the company’s performance in each category to its average performance across all products. The categories with the average customer rating most above their own average rating will be considered the best to shop for on that company’s website.

Data & Methodology

For this project, I scraped over 1,000 product ratings from Ikea, Wayfair, Ashley’s Furniture, and Haverty’s. In addition to scraping their product ratings, I also scraped information such as the number of reviews the product had, the product’s price, and the specific breakdown of the number of reviews of each type (1-star, 2-star, 3-star, etc) if available.

Once scraped, I cleaned the data using a variety of techniques in Tableau Prep and R. Once clean, I created all of the graphs above via code in R.

My name is Jason Liebmann and I am a senior at the Wharton School at the University of Pennsylvania studying Business Analytics and minoring in Data Science.

--

--