Rocket Companies drives 20% refinance pipeline growth using unified data on AWS

For decades, the homebuying process has felt disconnected. Customers search for a home in one place, secure financing in another, and manage servicing through a third. Rocket Companies recognized this fragmentation as both a customer pain point and a business opportunity. Its mission is simple: help everyone home. And that mission shapes every decision about technology, data, and customer service.

As Rocket expanded from digital mortgages to mobile engagement to AI-connected customer interactions, the volume of customer data grew exponentially. The company now serves 62 million monthly active users, 15 million servicing clients, and over 353,000 mortgage professionals—generating more than 25 petabytes of data. Every new channel and touchpoint generated more information about customer behavior and intent. But having data was not the same as understanding it.

Teams across the organization were developing customer views tailored to their specific needs, enabling innovation and faster delivery within their domains. At the same time, this created an opportunity to further unify definitions, pipelines, and business context to unlock an even more connected view of the customer. As Rocket expanded through acquisitions of Mr. Cooper and Redfin, integrating millions of new customer relationships became a catalyst for creating a more scalable and connected approach.

"To scale personalized customer experiences, modern applications and access to data wasn't just enough. What we required was a common understanding of our customer across the entire business value chain," said Garima Sharma, Vice President of Engineering at Rocket Companies.

Rocket recognized that conversion was no longer a funnel automation problem. It was a data and context problem. Modern customers expect interactions that are relevant, connected, and timely, whether they are browsing homes on Redfin, applying for a mortgage through Rocket, or managing their loan through Mr. Cooper. The opportunity was to build a single, unified understanding of the customer that every team could access and act on, turning fragmented touchpoints into a seamless journey.

Making that unified understanding usable at Rocket's scale required more than bringing data into one place. Every new AI or personalization use case could not afford its own pipeline, its own customer definition, and its own governance model. Rocket needed a repeatable pattern that any team, whether in search, financing, servicing, or customer engagement, could apply immediately: connect customer signals across interactions and systems, curate them into governed shared context, and activate that context wherever decisions, personalization, analytics, and AI happen.

The acquisitions of Redfin and Mr. Cooper made the opportunity even more compelling. With Redfin, Rocket Mortgage, and Mr. Cooper under one roof, Rocket now had the pieces to connect the entire homeownership lifecycle — find, finance, and own. But only if the data foundation could absorb new sources, new customer relationships, and new operational demands without breaking.

To operationalize this model, Rocket built its data foundation on four principles: one lake, one table format, one shared catalog, and one ingestion pattern. It was a commitment to reduce duplication, create shared context, and give every team the speed to move without rebuilding capabilities from scratch. “The architectural shift was treating customer context as a shared, governed product. Rocket’s architecture made that context reusable across teams, so new experiences could move faster without each team recreating its own customer definition,” said Sajjan AVS, Senior Solutions Architect, AWS Financial Services.

Rocket's architecture supports three phases of data flow: ingestion, transformation, and consumption, built on a standardized stack of AWS services. Data arrives through streaming, event-driven, and batch channels into Amazon Simple Storage Service (Amazon S3) using Apache Iceberg table format and Parquet file format. Amazon Kinesis handles real-time streaming ingestion, while Amazon Managed Service for Apache Flink processes streams at the ingestion layer. Orchestration and processing layers, powered by AWS Step Functions for workflow orchestration and AWS Glue for ETL and cataloging add business context as data moves through raw, processed, and conformed states. On the consumption side, Amazon Redshift serves data warehousing workloads, Amazon SageMaker AI supports machine learning model development and deployment, and the entire platform is governed centrally through AWS Lake Formation for fine-grained access control. This design eliminates one-off pipelines and gives every team immediate access to trusted, contextualized data without duplication.

From this foundation, Rocket created shared business definitions around customers, transactions, and operational activity. Every team uses the same context for activation. The most important capability this foundation produced was Client 360: a connected view of the customer across six dimensions, including preferences and behavior, demographics and household, Rocket relationships, and financial health. Client 360 became the basis for additional shared views: Transaction 360, Mortgage 360, and Lead 360. "We moved from duplicated effort to shared context, and from searching and connecting the data in silos to using data that's already aligned in business context," said Sharma.

The outcomes were significant and measurable. Through close partnership across data and technology, Rocket's refinance pipeline increased by 20 percent. The company achieved a 10 percent lift in conversion across customer journeys, banker follow-ups improved by 9 points, and recapture rates reached three times the industry average. These results were driven by a shared understanding of where customers were in their relationship with Rocket and the ability to engage them with relevant, timely experiences.

That same cross-functional foundation proved its resilience during the Mr. Cooper integration. Rocket onboarded 40,000 leads within 9 days, with the first loan closing in just three days, compared with an industry average of 30 or more days. This speed was made possible by years of investment across multiple teams to build a unified customer ecosystem rather than starting from scratch.

The foundation also unlocked AI at scale. Teams across the organization could rapidly build and deploy AI and machine learning solutions. Data scientists no longer spent months locating and stitching together information. They could focus on solving business problems. This shared ecosystem accelerated experimentation, improved model effectiveness, and now powers initiatives such as Rocket's AI-powered agentic pre-approvals, which are expected to increase lead conversion by 33 percent.

"Unified customer understanding is far more important than isolated AI initiatives," said Sharma. "AI will be fueled by that shared context of your business entities. Hence that consistent, common understanding matters more."

Rocket continues to build toward connecting search, financing, and servicing into a single experience for the customer. With its unified data foundation on AWS, the company is positioned to deliver on that vision at scale.

Three key lessons emerged from Rocket's journey:

1. Unified customer understanding matters more than isolated AI initiatives. AI is only as powerful as the shared context it can draw from.
2. Standardized data patterns create organizational speed at scale. When every team builds on the same foundation, integration timelines collapse, as the Mr. Cooper onboarding proved.
3. The value of data increases when it becomes operational across the business. Data locked in dashboards is insight; data flowing into real-time activation is revenue.