In its "US Digital Travel Sales" report, eMarketer says travel made up approximately 36% of all business-to-consumer e-commerce sales last year, but by 2016, that share will drop to less than 30% and continue to fall steadily.
In her blog, "Big Data and Digital Intelligence for Hospitality and Travel," Natalie Osborn, a senior solutions architect at SAS (this site's sponsor), explains the implications of this report: "It is no longer enough to have a website with a booking engine; the online experience must be such that it attracts customers and keeps them coming back."
How to do that is an issue that travel sites are grappling with as we speak. Kaggle, a predictive analytics firm known for its competitions, posted a "Personalize Expedia Hotel Searches" contest in September that drew some 337 competitors.
An Austria-based team from product recommendation software developer Commendo took the top spot, as announced today in a press release from Opera Solutions, a big data analytics company that purchased Commendo in 2012.
According to the release, Expedia offered $25,000 to the team that developed the most accurate predictive model for personalized search results -- results that customers were most likely to click and book. Each team received 2GB of shopping and purchase data on the individual customers and some 400,000 customer queries to resolve based on 50 variables (such as smoking, location, pet-friendly, and so on).
In a blog about the contest, Opera Solutions senior editor Sarah Anderson says this model wasn't a simple matter of determining which results match all the customer queries, but of predicting which should be ranked highest:
The one listed first is most likely to be clicked on and ultimately purchased. And seeing as how we humans don't have long attention spans, travel agencies only have a few seconds or a couple of clicks before the user decides the responses aren't up to par and leaves the site altogether. So figuring out which one should come first is the question our scientists and hundreds of others needed to answer.
For the analytics professionals in the audience, Normalized Discounted Cumulative Gain is the metric used to determine the optimal usefulness of each result -- rewarding competitors for surfacing the best results highest on the list.
Machine learning for travel booking plumbs a customer's click history to determine whether he or she has children, prefers bathtubs to shower stalls, or orders from room service often, Anderson says. Some of these may indicate a stronger preference for hotels with pools; others may indicate a preference for hotels with fitness centers. The magic lies in putting all the customer attributes and hotel attributes together with the query (number of travelers, time of travel, number of rooms requested, and so on) to predict the most likely choice.
"Even if you don't realize it at the time, you're inputting those subtleties every time you search for a hotel room -- even if you don't make a purchase," Anderson writes. And the ability to predict accurately is "pure gold" for online travel companies.
What do you think, members? Would a site that seems to know your preferences win your repeat business? Or is price the ultimate factor whenever you're planning a trip? Share your online travel experiences in the comments.
— Michael Steinhart, , Executive Editor, AllAnalytics.com