Travel technologists press Saber into an agreement with Google • The Register

Feature The computing and travel industries have walked hand in hand for decades. For perspective, American Airlines signed a deal with IBM in 1957 that developed the first computer reservation system in 1960, based on two IBM 7090 computers.

Over time, the reservation system itself has transformed into one of the three largest global distribution systems (GDS), called Saber, which trades in flights from many airlines, providing them to travel agents and consumers.

But with this heritage comes a problem when it comes to serving consumers in the 21st century.

“The data formats in the travel industry have been around for a very long time; for decades,” says Andrew Gasparovic, chief architect at Saber Labs, GDS technology division, which was founded in 1996.

“They were designed at a time when there wasn’t a lot of thinking about what you can do with data. They weren’t amenable to what we call the concept of ‘display and order,'” he said. record.

Fetching data about customers’ last offers and reservations is common in e-commerce and helps sellers anticipate what customers are likely to buy next.

But according to Gasparovic, bringing this data together across more than 12 billion shopping requests and one billion travelers each year was no trivial task and was tasked with a partnership with Google formed in 2020.

Saber is in the process of moving its IT infrastructure to Google Cloud. It also adopts operational data tools including Spanner managed systems, a distributed database that supports GoogleAds and BigTable, and a NoSQL wide column and key value database.

In analytics, you use BigQuery, Google’s distributed data warehouse.

Other GDS systems include Amadeus GDS—created by Air France, Lufthansa, Iberia and SAS in 1987 as a European alternative to Saber—and UK-based Travelport (which includes Apollo, Worldspan and Galileo GDS.) All of these networks began managing life tickets between Airlines and travel agents, but today they also work with travel websites, car rental companies, and hotels.

Reservation please?

Saber’s oldest travel distribution network predates the Internet. The airline’s reservation database stores passenger reservation information, seat selection, tickets, special requests, and other important information about their flight. Saber typically processes thousands of reservation updates per second on behalf of carrier customers. The airline’s reservation database must be provided from many areas of availability.

Meanwhile, the flight shopping system creates millions of routes per second on behalf of travelers using mobile apps, third-party travel sites and airline call centers. Manages 10 exabytes using Bigtable. Building a data solution to anticipate what customers might buy next is built on top of an existing data warehouse, which includes Teradata, Oracle, and IBM.

“We have pretty much everything you can imagine in terms of data warehousing and operational data warehousing technologies,” Gasparovic says. “For all of these existing systems, we’re thinking about how to get a feed first in BigQuery.”

“It was hard to do things like understand the traveler as they went from shopping for flights to booking flights. To use this information to understand these travelers’ interests, what their preferences are, and what kind of product packages they usually buy together it was hard to do that kind of thing because this data They exist in many different systems in many different forms.”

Saber Labs keeps what was submitted to the traveler, and creates a view ID that flows into BigQuery along with what the customer requested. “It’s not just about getting it in the same place,” he says. “It’s updating and updating the same data model to be able to identify those things with a unique key.”

Being able to bring it all into the same system, and begin to relate and understand it as a whole was the first step. Next was building machine learning models that learn from data. “This is a really big deal for what we can offer our customers,” Gasparovic says.

But machine learning comes with a caveat. It is possible to manipulate machine learning forever and not necessarily solve a real problem. He says projects should start from practical problems to solve.

Saber Labs’ approach is to compress the training phase of an ML project by using reinforcement learning technology to try something very quickly and see what benefit it provides. “You can think of sort of like a very cool kind of AB test with many different items being tested at the same time,” Gasparovic says.

“The nice thing about it is that we don’t have to do a lot of pre-training the models and trying to manipulate things to get accuracy. And when we know it really does offer some benefit, we can go back and do some classic machine learning: building the models and doing the supervised learning process where you train a model on The data and then put this model into production and use its output to make predictions, but starting with this experiment was really, really helpful. We have to know where we want to spend our time.”

It has already introduced machine learning models that help its partner organization deliver better offers to travel customers.

“When we’re able to put that on a website and actually see the impact, it’s something that customers have come to expect, but it’s been very successful for hotels and airlines. And enabling this for them is still something unique in the industry,” Gasparovic says.

The Google Cloud solution is built on Saber’s existing data repositories, but in the long run the plan is to consolidate the data into a single system, but for some time, Saber will likely power both worlds with the new, multi-layered world on top of the existing one, he says. .

In the meantime, Gasparovic expects more victories to come from the current group. “We just scratched the surface of what we can do with machine learning and experimentation,” Gasparovic says. ®

Leave a Comment