TransUnion transforms its business model with IT

Count TransUnion among the rising tide of enterprises evolving their identities thanks to IT.

“We are thinking like a software company and transforming ourselves like a software company,” says Venkat Achanta, chief technology, data, and analytics officer of the $4 billion credit bureau, which is recasting itself into a customer data services provider intent on parlaying its reputation for trust and ample data assets to drive analytics, machine learning (ML), and AI development on the cloud.

The information and insights company’s foundation remains ensuring that every consumer is accurately represented in the market. But following its $3.1 billion acquisition of data and analytics company Neustar in 2021, TransUnion has expanded into other services such as marketing, fraud detection and prevention, and robust analytical services.

At the core of its strategy is the mountain of data that TransUnion has acquired — along with more than 25 companies — over decades. That data is in the process of being unified on a multilayered platform that offers a variety of data services, including data ingestion, data management, data governance, and data security.

Once completed within two years, the platform, OneTru, will give TransUnion and its customers access to TransUnion’s behemoth trove of consumer data to fuel next-generation analytical services, machine learning models and generative AI applications, says Achanta, who is driving the effort, and held similar posts at Neustar and Walmart.

The power of productizing data

TransUnion’s OneTru has been made possible by the company’s migration to AWS, dubbed “Project Rise,” which is slated for completion by year’s end. Following its acquisition of Neustar, a Google Cloud Platform customer, TransUnion embraced a multicloud infrastructure that also supports GCP, but the crown jewel of its technology modernization is OneTru, and its 50 petabytes of data assets amassed over decades.

While Achanta concedes it’s possible to deliver such services using standard data lakes and data management functionality from cloud providers and SaaS vendors, TransUnion’s proprietary cloud-native data platform will deliver high-end services with added assurance that the data sets are uniform and consistent, he says.

“I need a consistent platform for data ingestion, how I think about data management, data governance, and how we think about [AI] model deployment,” says Achanta, whose transformation relies on thousands of engineers and more than 700 data scientists across the organization. “We’re modernizing existing products to get to this entire data analytics value chain.”

The multilayered data platform will enable TransUnion’s customers to perform deep analytics and build complex AI models. Under the hood, OneTru’s data management layer, identity layer, analysis layer, and delivery layer leverage a robust data governance framework and standard access control to ensure legal and regulatory compliance and auditability, Achanta says.

For example, in every jurisdiction where TransUnion operates, OneTru’s architecture has multiple mechanisms for legally managing the separation of data, including physical separation through separate instances, to logical separation via workspaces, data access groups, and user permissions, according to the company.

The proprietary analytics engine drives significant internal efficiencies and gives the company new opportunities for delivering AI-powered data services. It also enables customers to tap into that mountain of data and apply advanced AI applications to mine more value out of it.

TransUnion’s strategy provides all the important services enterprise customers need to build traditional AI and generative AI models, including storage, identity resolution, fraud prevention, and governance, particularly for those in the financial services markets, according to one industry analyst.

“AI models are only as good as the data that’s being fed into them and being used to customize them,” says Arun Chandrasekaran, an AI analyst at Gartner. “Data has gravity. It is expensive to move data and there are these stringent regulations and privacy considerations around moving data as well. Eventually, AI models will get slimmer and will move closer to the data.”

The platform approach to AI

TransUnion has been developing, deploying, and continuously modifying machine learning models for some time. It is the “bread and butter” of the business because deep ML helps TransUnion “understand the grammar and pattern of fraud, and fraudsters change every day,” Achanta says of the company’s “adaptive” machine learning models for fraud and credit scoring.

This adaptive capability enables data scientists to modify ML models for precision and stability, and they are recalibrated daily to keep up with bad actors, he says. Credit score models are also constantly being modified.

IDC analyst Sean O’Malley says a number of companies besides TransUnion are developing AI engines on data platforms to extend their data and analytics services to enterprise customers, such as Experian and Visa.

To date, TransUnion has relied on more than 15 different products for data ingestion, data management, data governance, and data security, all of which take data out of silos to make it available for use, Achanta says. But the resulting OneTru aims to provide a uniform, consistent data platform to ensure accuracy and deliver greater efficiencies for internal uses — and perhaps most important, for customers who want access to that goldmine.

It only “make sense for our customers to be able to take action on the data,” Achanta says, adding that TransUnion owns all the core IP that came with Neustar, and its products can be deployed on multiple clouds.

“That is the beauty of the platform,” he says. “We may be able to analyze data quickly, develop a model and deploy it connected to a real-time API quickly. We can do all that in weeks now instead of six months.”

As for generative AI, TransUnion has started to experiment but is not yet in production. As part of its testing, employees have access to conversational agents built using three different large language models (LLMs).

OneTru’s AI- and ML-based knowledge graph features, for example, will enable customers to break down data previously stored in siloes and apply generative AI to move beyond “traditional identity graphs that are limited to explicit data and deterministic linking,” according to the company.

Achanta says customers will have access to knowledge graphs that retrieve information from a wider field and will be empowered to make logical inferences from linked, unstructured and structured data, enhancing the meaning and context. For instance, OneTru’s AI graphing capabilities will improve identity resolution in fraud cases.