Orbitz: Mac users book fancier hotels than PC users

By Barbara De Lollis, USA TODAY

Updated 5/9/2012 5:26 PM

Readers: This is the third post from Orbitz Worldwide CEO Barney Harford, who is guest blogging this week. Here's Harford, in his own words:

By Jack Gruber, USA TODAY

Much of what we do behind the scenes at Orbitz centers on the way we handle data. Every day consumers execute millions of searches on our websites around the world, and we have invested significantly in being able to log information about these searches to learn more about the travel planning preferences of Orbitz users. By analyzing variations in different user and session attributes, we can make more personalized recommendations to users conducting searches on our sites in the future.

We're trying to capture more information about travel planning behavior than any other company in the world. To give you a sense of how much data we're talking about, last year we processed approximately 750 terabytes of log file information about user sessions. Just one terabyte can store the information in 285 million pages of text.

Here are examples of some of the initial insights we're uncovering, and how we're using them to improve the experience for travelers shopping for hotels on Orbitz:

We've identified that Mac users are 40% more likely to book a four- or five-star hotel than PC users. A similar skew applies for iPad users. We can use that information to influence which hotels we recommend to users we see searching on a Mac or an iPad versus a PC for example. On our website, once you get to the page for a particular property (let's call it "Hotel A") we show consumers a list of alternative hotel recommendations. This list is primarily made up of nearby properties that were ultimately booked by customers who had also viewed Hotel A. That's a pretty useful feature already, but we're then able to personalize that list by taking into account factors such as whether we see that the user is using a Mac or a PC.

Similarly, if you start a hotel search and tell us you want to visit Orlando this summer with your kids, you'd probably hope to see on that first page of results a list of hotels that include options like a swimming pool, rooms with two beds and free breakfast. On the other hand if you are a romantic couple traveling without kids, you're likely going to want a hotel that has a more stylish feel, and potentially one that specifically doesn't cater to families. We're able to look at the way a given hotel's conversion varies based on whether or not the traveling party includes kids or not. Using this information we can create a "kid friendliness" score for hotels that are particularly preferred by people traveling with kids; similarly we're able to create a "kid avoidance" score for hotels that adults traveling without kids seem to avoid! Those scores are then used by the algorithms that decide which hotels to put at the top of the list of results we show to customers when they're searching on our site, based on whether we see them traveling with or without kids.

These are just a few examples that we're getting started with, but there's so much more opportunity for us to use techniques like this to deliver a more and more relevant set of options for our customers. We know that 90 percent of customers book a hotel from the results they see on the first page of results (typically the first 25 hotels). Furthermore we know that 50 percent of customers book one of the top five properties we show, and a remarkable 25% book the top sorted property. If we're not showing relevant options to our customers, they will look elsewhere to make their selection. Getting a highly relevant set of options into those first five slots is really critical for us, and we think there is huge opportunity here for us to use personalization to improve the trip planning experience for our customers.