Travel Has Much to Learn from E-Commerce – Recommendation Engines
In 1999 I had an opportunity to work with a variety of best of breed e-commerce products for a large retailer in the U.S. midwest. One of the products I worked with was NetPerceptions. The software was developed out of a research program at the University of Minnesota in 1992 and is based on the concept of “collaborative filtering”. Essentially the software analyzes patterns in purchase history and then makes recommendations based on what other people buy. This software is the basis of Amazon.com’s widely popular “People who bought this also bought:” engine. In the broadest sense, the software is a recommendation engine that provides the customer with choices based on the buying patterns of thousands, if not hundreds of thousands of consumers before them. For example, if you are looking at purple towels, the system will look for products that were purchased by people who also purchased purple towels and then make the recommendation to you. It seems pretty straightforward, but the underlying technology is very sophisticated and requires a tremendous amount of product data and purchase history. The technology has also been around for over seventeen years!
Given that this is not new technology, why have we not seen any “real” recommendation engines in the travel space? Simple, in order to make recommendations you have to be able to clearly understand what you are selling, and in my opinion, most travel suppliers don’t know what they are selling. When a customer is booking a flight to Las Vegas, or a room at a hotel in Las Vegas, it is already too late to be using a recommendation engine. I realize hoteliers are probably not very happy reading this, but the bottom line is that the only true product you can sell is a destination. Everything else is simply a means to an end or an add-on.
For example, if you looked at booking history as a basis for recommendations, what would the recommendation engine suggest if you booked a ticket from Vancouver to Amsterdam? “People who booked flights to Amsterdam also booked flights to Berlin, would you like to book a flight to Berlin now?” Not very useful I think. Or how about the hotel example “People who booked the Fairmont Waterfront, also booked…” what? another hotel at the same destination? I don’t think so. But… when you apply the recommendation engine on the destination level, now you have a true basis for comparison. For example, “People who travel to Mexico also travel to Cuba”, or perhaps even more interesting when applied against activity choices “People who enjoy scuba diving travel to Bermuda” now you have a powerful and compelling tool for helping travelers make decisions.
So what can we do to help make recommendation engines a reality? Firstly, we need to create a genome for destinations. Pandora did this for music and it has taken the independent music world by storm. The same could be true for travel, if we can classify, tag, and standardize a genetic code for every destination in the World, we would have an incredibly rich database filled with destinations that could be compared and contrasted based on a known set of criteria. This is where a standards body like the OpenTravel Alliance could play a major role in defining how destination data is used in travel product transactions. Once the genome is created, we would need every destination in the World to agree to populate and maintain this data in an accurate manner and share it openly for the purposes of improving tourism Globally. Once the content is standardized and populated, the recommendation engines will come and the landscape of travel booking will change forever.
A dream? Certainly, but we have to start somewhere.
