When was the last time you booked a hotel room? Chances are you probably used an Online Travel Agency (OTA) app like Expedia on your smart phone or went on the TripAdvisor website first to check out the hotel. The last time you purchased a product from Amazon, you probably checked out how it was rated by others. And, if you read the reviews, you are not alone. On an average, users read at least seven reviews on TripAdvisor before they make their hotel bookings. By the way, when you purchase tickets for movie, do you check out how others rate it on Fandango?
Historically hotels would take into account competitor pricing, customer willingness to pay, inventory, and forecasted demand to determine the optimal price to charge for a room. While price was and will continue to be a significant lever in purchase decisions, user ratings offer an extra factor to help understand the booking behavior of hotel guests.
Crowd Sourcing for a product’s quality has significantly altered the online buying landscape, be it for hotels, retail, or movies. It has significantly changed how people are making their purchase decisions. It is not just consumers who can benefit from it. Hoteliers can use online reviews and ratings to price their product optimally. Just like how customers can choose not to book a lower rated hotel, a higher rated hotel can leverage these ratings to increase their prices. Think of it this way: would you pay 25 cents more for a Jason Bourne movie which is rated 4-stars on Fandango vs. another rated 2-star movie running at the same time?
What are user ratings?
User ratings are the numerical aggregation scores normally on a scale of 1 to 5 across multiple channels like TripAdvisor, Expedia, Priceline etc. Whenever a site visitor posts a review about his or her experience at a hotel, it is combined with other scores to come up with an average user rating score for that hotel. While TripAdvisor is considered by many to be the most trusted channel, various brand.com channels scores are equally influential. Some OTAs actually show both their scores as well as TripAdvisor scores to be transparent.
Why are user ratings important?
There is enough empirical evidence which suggests that guests perceive user ratings as an extremely important factor while making their booking decisions. Research by Professor Chris Anderson at Cornell University has been widely cited in both academic and industry circles. According to his research, an average one-point increase in user ratings scores can help hotels justify an 11% price increase without giving away the market share. The hospitality world certainly understands the importance of an 11% yield. This is just too crucial to underestimate.
Challenges with incorporating user ratings
User ratings eventually converge to a particular band as they reach a critical mass. For example, a hotel in Dubai is rated on average as 4.5 out of 5 stars on TripAdvisor. But, as data is collected, it eventually plateaus and this is a primary point any statistician has to take into account. Consider also that a rating of 4.5 stars from just 5 reviews is not that powerful as a rating of 4 stars from 500 reviews. The number of reviews that went into the overall score, is a good indication of the reliability of the ratings score. Speaking of reliability, Choice hotels requires you to have the PNR number of your booking to ensure data sanctity and Amazon certifies a rating as a verified purchase.
On top of numerical scores, there is plenty of unstructured data to be analyzed in the form of user comments, pictures, etc. This requires machine learning tools like Natural Language Processing, Text Miner etc. and has significant bearing on the overall sentiment of that hotel.
JDA has the solution
During conversations with top hotel chains who also happen to be JDA customers, we found they acknowledged that ratings are important, but they were not sure how to monetize or quantify that aspect of ratings. For example, a top hotel chain is aware of their user ratings but were not sure by what percentage should a favorable rating result into a price increase and vice versa. Without analytics, it is difficult to predict how much a ratings difference will affect their natural demand. The million-dollar question is what price should I charge for my hotel rooms given my user ratings and that of my competitors?
JDA’s Pricing and Revenue Management (PRM) Product Management Group recently partnered with RateGain to gather user ratings data. They also worked with data scientists from JDA Labs who designed a unique algorithm (patent pending), which lets hoteliers quantify the impact of user ratings on their hotel room pricing decisions. We collected the data for more than a year and married the past hotel pricing with user ratings at a given point in time. This was fed to our algorithm which helped us determine the impact of ratings and subsequently churn out an optimal price. The objective of our algorithm was to maximize revenue and we found that anytime a hotel was rated better, our model came up with increased revenue than what was actually charged. Conversely, when the hotel was rated lower compared to the competitor set, we did see evidence of revenue drop. This way it becomes a leading indicator instead of lagging indicator.
We not only have filed a provisional patent on this approach but we just validated the prototype with actual customer data on sample properties. The results are pretty exciting so far and we are in process of testing them further.
For one sample property we found our model suggesting 13% increase in revenue keeping everything else the same because the property was rated much better than its competition. Even if we were 20% accurate in our analysis, this translates into 2.6% revenue improvement. Translate that across millions of dollars in annual revenues of major hospitality chains and you can clearly see how this can be a game changer. Best part, hotel chains almost have everything they need to implement this. They just need to have a partner like JDA to unlock this potential!
Applications in retail
While we primarily started this work with hospitality in mind, we do see significant applications across retail and possibly other verticals.
For example, while optimizing assortment, a part of the decision is what products should I add or drop from the assortment. Adds are harder because you don’t have any sales history. Retailers normally try to get market data from Nielsen which can be expensive. Something of value and less expensive can be ratings data from Amazon. A retailer can easily incorporate ratings data to anticipate any corresponding pickup or drop in demand similar to a hotelier. The opportunities are endless.
In the end, as an avid Bourne fan, I would gladly pay that extra 25 cents for the Jason Bourne movie!