Priceline: Helping travelers curate unique experiences by improving IT data management

In the early days of the internet, Priceline founder Jay Walker had a radical idea: what if customers could name their own price on partially-empty flights or unused hotel rooms? One airline executive told him they’d be better off if he were "hit by a bus" than embrace his idea. Despite the resistance to his pitch, Walker knew he had a winning idea and one that relied entirely on the power of the internet to come to fruition.

Walker pushed past the resistance to build Priceline and help people experience the moments that matter by offering them the best possible deals. As an ecommerce pioneer, Priceline understood the need for innovation, but as the company grew, its IT infrastructure aged. Its legacy code base was spread across three disparate data centers, making it more difficult for product and engineering teams to keep up with the company’s growth goals. Multiple teams had to coordinate schedules for deployments and A/B testing, and these workflows slowed the time to market for new products and features.

​​Running this rigid, legacy infrastructure was also capital intensive, and teams would need to plan four or five months in advance to allocate resources. "Increasing the productivity and velocity of our product and engineering teams was paramount," says Marty Brodbeck, CTO of Priceline. "We needed to be more flexible for customers and make it easier for them to search, price, and book faster." The team worked with Google Cloud to create an agile technology platform that could fuel its growth.

Migrating a global IT infrastructure to the cloud to help Priceline scale

The company's vision required a fast, reliable tech stack that could support Priceline's operations worldwide. It also needed a scalable infrastructure that could automatically scale up and down based on customer demand, which brought Priceline to test Google Kubernetes Engine (GKE).

Priceline piloted Google Cloud on its rental car stack. After the success of this program, the company decided to accelerate its cloud transformation and migrated the entire product platform to Google Cloud in less than two years. "Other cloud vendors couldn’t compare with the reliability and computing capabilities we saw with Google Cloud. They were black boxes. We didn't know what was wrong if something stopped working," says Brodbeck. Coinciding with its cloud migration, Priceline shifted to DataStax Astra DB, a managed services approach for its Cassandra database that supports its booking and checkout flows which communicate directly with GKE. This interconnected, flexible infrastructure allows Priceline to better manage, analyze, and act on its datasets.

Meanwhile, with BigQuery, the company can normalize thousands of global data sources, including customer data. This way, Priceline can review patterns in its data storage and processing usage to make more informed decisions about resource allocation, product performance, and more. This includes near real-time anomaly detection so teams can make adjustments and prevent additional surprise costs related to data storage and processing. With the data collected and monitored with BigQuery, Priceline can rely on Dataflow and Dataproc for processing, BigLake for unifying data lakes and warehouses, and Dataplex for security and governance to keep its data safe.

"I view Google Cloud as an extension of our engineering team. The less time we have to manage and maintain data infrastructure, the more we can focus on core capabilities," says Brodbeck. "Having these highly scalable, SaaS-based data platforms means we can focus on our customers."

The changes to Priceline’s infrastructure helps the company provide the best deals to customers as efficiently as possible. This involved building Priceline’s new customer loyalty program, Priceline VIP, on BigQuery to help customers unlock better deals. Vertex AI powers many of the platform’s customer-facing features, such as improved sort algorithms, pricing personalization, and rewards management.

Reinventing the customer travel experience using more-informed, structured datasets

With more efficient reporting and the ability to make decisions faster, Priceline is empowered to react and adapt to market dynamics more quickly. For example, during COVID-19 lockdowns, it was able to create a new product to meet a nearly overnight surge in demand for road trips when customers would rent a car for a destination trip rather than flying. "The developer pipeline that we've built on top of Google Cloud, Plaza, has enabled us to automate the design, development, and deployment processes," says Brodbeck. This shift has more than doubled the organization’s speed of executing new deployments.

As COVID-19 lockdowns lifted, the company saw a dramatic increase in travel beyond road trips as flights and longer vacations roared back. Its new, flexible infrastructure meant Priceline could easily scale up based on the increased customer traffic without having to track down new server hardware. The power to scale usage up and down as needed continues to give Priceline the ability to adapt to meet customer demands during peak travel times.

Using BigQuery, the team has been able to review data, including these traffic trends, faster than ever. Priceline can identify almost real-time anomalies in data storage usage that would’ve otherwise only been caught at the end of a month or quarter.

From a production perspective, the organization can also find ways to improve its development process.

"Google Cloud Storage Insights has delivered significant value, offering usage patterns that inform real-time, data-driven decisions for business improvement. It facilitates rapid creation of custom alerts, notifications, and dashboards for informed storage optimization, saving weeks of development time" says Sachin Gopalakrishna Menon, Head of Data and Marketing Technology, Priceline.

Designing the ideal trip with an AI-powered platform

Next, Priceline is exploring ways to create a real-time data infrastructure to build dynamic features based on events in Apache Pulsar in combination with Astra DB on Google Cloud for its vector search capabilities. Astra DB's vector search for Generative AI provides the high performance, availability, and scalability Priceline needs for real-time data. The company plans to use these platforms together to automatically add new features to machine learning models and create a scalable feedback loop.

Priceline is also working closely with Google Cloud to further enhance its customer experience. The company is incorporating generative AI to provide points of interest to customers booking hotel stays. "When I was little, my mom would call a travel agent and spend hours on the phone planning out a vacation for three little boys. Working with Google Cloud to incorporate generative AI allows us to create a bespoke travel concierge within our chatbot," says Brodbeck. "We want to help our customers go beyond planning a trip and help them curate their unique travel experience."