How Expedia is consolidating SaaS applications and managing 70 petabytes of data

While most people are familiar with Expedia as a place to compare flight prices and showcase Ewan McGregor's talents, behind the scenes the company is both a consumer destination and a business-to-business SaaS provider. Rajesh Naidu is responsible for maintaining a tech infrastructure platform that can meet the needs of both lines of business.

"It's an interesting model. We are serving not just our own base, but also our B2B partners," Naidu, senior vice president and chief architect at Expedia, said in a recent interview. "That makes us unique in that regard of consuming SaaS as well as providing a SaaS service."

While that interview was conducted before Ariane Gorin, formerly the president of Expedia for Business, was named CEO of the whole company last week, it's not hard to imagine with her appointment that balancing the seasonality of travel booking with the resilience demanded by enterprise customers will continue to be a tech priority for the company. Expedia sells its technology to enterprises looking for an internal travel concierge as well as third parties that want to use its APIs to create their own travel-booking experiences on their sites.

"You can create differentiated front-end pieces, (but) have some similar pipelines in the back to connect to supply," Naidu said.

Expedia has been a predominantly AWS shop since 2012, but like many companies with a market cap around $18 billion, has a "small presence" (as Naidu put it) with other cloud providers and operates a data center simply to run Oracle's ERP software.

And like many companies that have grown through acquisitions over the years, one of Naidu's jobs has been to integrate those acquisitions onto a common tech stack, which requires taking a hard look at the SaaS applications used by those companies and deciding whether or not it makes sense to bring them forward.

"We are always looking at minimizing the footprint in terms of not proliferating five different SaaS providers for the same capabilities, because we were in that situation as we acquired these brands," Naidu said. For example, Expedia is currently looking at consolidating its use of Salesforce, which it is otherwise happy to keep using, just not in redundant ways that can be centralized.

Those acquisitions also required Expedia's tech teams to consolidate a lot of different policies and services related to data. "We have more than 70 petabytes of data that we have acquired over the years," Naidu said, but Expedia separates consumer data from the B2B data and therefore needed to architect what he called "the right guardrails" to prevent that data from mixing while still being able to leverage it for analysis on the back end as it integrated those companies.

To that effect, the company recently figured out how to streamline its machine learning platforms, turning nine different, separate platforms into a single one across the company. "I think the big shift around data has been being able to serve up the right data, being able to govern and measure the data, (and) do rapid experimentation," Naidu said, and it's much easier to accomplish those goals when data isn't scattered across the company.

Expedia was already a big proponent of observability technology across its infrastructure, but is starting to invest in data observability as well. Juggling data from hundreds of different hotels, airlines, and car rental agencies is an extremely important part of the company's business, and there are obvious ramifications if the wrong data is in the wrong place.

If everything else is running like a well-oiled machine, it's then going to come down to the pipeline and whether the information is flowing from point A to point B.

When confronted by a slowdown in bookings in the past, Expedia had generally assumed that something was wrong with its infrastructure that was preventing customers from executing a transaction. But it realized through data observability that people might not be booking that hotel room in Las Vegas because the hotel sent Expedia the wrong data, and the room was being sold for $1,000 a night when it should have been going for $100 a night, Naidu said.

"If everything else is running like a well-oiled machine, it's then going to come down to the pipeline and whether the information is flowing from point A to point B," he said.

And like most companies in 2024, Expedia is turning its attention to generative AI and the possibilities it could unlock for its business.

The company is using ChatGPT on the consumer side of its business to help customers book travel through a text conversation, and it also launched an internal pilot of GitHub Copilot that one-third of its developers are currently using. "We're looking at generation of unit test cases, code refinement, improving productivity and pushing new code every three days into production." Naidu said. "We're starting to see some good productivity gains."

But as a big Microsoft Office shop, Expedia is waiting to see how Microsoft Copilot for 365 evolves before shelling out for its new generative AI features, which have seen mixed reviews in the first few months they've been available.

And when it comes to another tech product that has seen mixed reviews in its early days, Naidu is personally excited about the potential for developing Expedia applications for Apple's Vision Pro and how that could transform the travel-booking experience.

"I wonder what use cases there could be for travel down the road with that, right? Especially from a discoverability perspective; what if you really want to go see what a location looks like?" he said.