BY Capgemini
This page was produced by FT2, the advertising department of the Financial Times. The news and editorial staff of the Financial Times had no role in its preparation.


IT trends spotted and checked by experts

Chief technology officer at KAYAK

Mr Giorgos Zacharia is chief technology officer at travel search company KAYAK, after previously serving as Chief Scientist. Giorgos has extensive experience in machine learning and data mining...


Transforming customer experience with superfast analytics

26 May, 2014 08:07 am

We process over a billion queries per year, from travellers looking for flights, hotels, vacation packages and car hire. In order to deliver the right experience, we rely on superfast data analytics.

KAYAK is a technology company, as much as it is a travel search provider. 70 per cent of our staff are technology employees, ranging from engineers to data scientists and those maintaining technical operations.

We face several complex challenges in our business. The first is accessing and analysing a huge amount of data from hundreds of airlines, hotels and travel agencies, and doing so as quickly as possible.

Useful analytics at speed

Some of the companies we work with are much faster than others at presenting results. In order to offer results quickly and usefully to our customers, we have a live stream in which we quickly show them the main results to begin with, adding additional items as they arrive during the following seconds.

For the companies that are particularly slow at providing information, we draw information from global distribution systems instead, in order to more quickly provide the data to users.

A major issue in the travel search industry is accuracy of data - when people search for flight or hotel prices, they can find that the price at the point of booking is different from the original price presented on the search engine. Big data analytics enable us to check the problematic searches and companies involved, so that we only present the most reliable information.

Advanced data mining also helps us know the final prices on offer, including booking fees. Many search engines provide basic prices that do not include additional booking or credit card fees. When a customer searches for travel prices, we dig deeper to mine the actual costs of their preferred payment method, which they can select   , and show the final cost as closely as possible in our payment calculator.

The take-off of predictive analytics

We offer predictive analytics directly to users. Since January last year, we have offered users the opportunity to see how their flights prices are likely to change in the coming weeks (see screenshot below).

In order to do this, we rely on a number of items of data, including historical price information, an analysis of repeatable patterns, and monitoring changing current demand.

We present predicted price changes in a simple graph format, including information showing how accurate we expect this prediction to be, so that users easily know how and when to act. Since we first trialled this option, we have seen many users accessing the information to hold back on purchases until a time when there is a better deal, or buy right away when we forecast the prices to go higher.

Presenting complex data

KAYAK users access our service on a number of devices, ranging from traditional PCs to tablets and mobile phones. This presents us with a real design challenge, given that we collate and show so much information to each user in one display.

On PCs and tablets alone, with their larger screen size, there is the challenge of showing the key information such as price and times, as well as filtering options, in a simple and easy-to-use fashion. Presenting this on mobile is even more complex, but the tendency of users to scroll does help us provide further filtering options lower down on the screen.

Technology fueling the change

In addition to our own extensive work in improving analytics, it is also the technology development by the industry that has enabled us to drive the change.

While some of the analytical technology, including machine learning, has been effective for much of the last decade, it is the real data mining that has transformed. We are now able, using technology such as Hadoop, to dive much more quickly into the data and arrive at what we need.

These developments also help us cope with fluctuations in traffic according to seasons and promotions. We always over-plan for traffic, and have two data centres ready to handle the capacity. By keeping excess capability, we are also able to switch off elements of either datacentre for maintenance, without impact.

On the horizon

We are finding that mobile search is growing significantly, and represents many searches by business travellers and those already on holiday, looking to change plans. Tablet searches are particularly popular at night when people are at home making holiday bookings.

New technology is always emerging to support these changes. We will continue our advanced work with analytics, so that we can always provide the fastest and most accurate holiday results in the marketplace.