Using data to balance fraud prevention and user access at scale for Login.gov

About GSA

The General Services Administration (GSA) is a federal agency responsible for supporting the basic functioning of other federal agencies by providing centralized procurement services, real estate management, technology solutions, and administrative support across the government.

Login.gov serves as a critical piece of digital infrastructure for the federal government, aiming to become “one account for government.” The service is a single, secure, and accessible identity verification and authentication platform that agencies can rely on and users can trust.

The challenge

With over 100 million user accounts and more than 30 million people who have completed identity verification, Login.gov already demonstrates significant value. 25% of verified users have connected their account to two or more government applications, showcasing the network effect of reusable identity proofing. With the recent release of OMB Memorandum M-26-18 directing agencies to adopt Login.gov as an identity verification option, demand on the platform is expected to accelerate rapidly.

This growth surfaces a persistent, high-stakes tension at the heart of Login.gov’s mission: balancing fraud prevention with user access. Every control introduced to stop bad actors risks adding friction for legitimate users, especially those who are not tech-native, distrustful of digital institutions, or lack conventional forms of ID. Because Login.gov handles highly sensitive data like Social Security numbers, selfies, and phone numbers, it must maintain rigorous security and compliance standards. Any changes to how sensitive data is handled and stored requires extensive security review, and improvements must be deliberate, well-evidenced, and carefully validated.

Login.gov has worked to ease this tension by building optionality into its proofing flow: remote or in-person verification at the post office, multiple accepted ID types, varied customer support channels, and newer options like Mobile Driver’s Licenses and agent-assisted proofing through a pilot with Social Security Administration (SSA) field offices. But optionality alone isn’t enough—the team also needs rigorous, timely insight into where users get stuck, whether fraud controls are disproportionately burdening legitimate users, and how new pathways are actually performing, all without weakening the platform’s security posture.

This is where trustworthy data becomes mission-critical. Without clear visibility into user behavior, drop-off points, and fraud patterns, product and policy teams risk making consequential decisions—ones that affect whether someone can access government services and benefits—based on incomplete information. And the stakes are rising: as M-26-18 drives more agencies and users to Login.gov, the costs of both excess friction and unchecked fraud scale with it. Getting the data pipeline, modeling, and reporting right isn’t a nice-to-have; it’s foundational to fulfilling Login.gov’s mission responsibly at scale.

The approach

As a part of Login.gov’s Data Team, a U.S. Digital Corps Data Analytics Fellow worked to transform raw application logs into organized, accessible data that can directly inform product, fraud, and partner-facing decisions. By contributing to the three key interconnected areas—data engineering, data modeling, and reporting—the Fellow helped close gaps between what the platform was doing and what stakeholders could actually see and act on.

The Fellow:

• Diagnosed friction points across the proofing funnel: The Fellow analyzed where and when users got stuck across Login.gov’s broader identity verification flow, surfacing patterns tied to ID types, verification methods, and user segments. This fed directly into Login.gov’s roadmap for new pathways to help more users “get to yes” without loosening fraud controls.
• Evaluated new proofing pathways with real data: During Login.gov’s new agent-assisted ID proofing flow pilot at Social Security Administration (SSA) field offices in the fall of 2026, the Fellow, once he was placed full-time on the team, analyzed where users got stuck and what distinguished successful completions from drop-offs. This work relied on the Fellow’s deep understanding of the business logic and data modeling pipeline that was built during his time with USDC. These findings went directly to SSA as actionable feedback, shaping decisions about how to refine and scale the pathway.
• Strengthen Login.gov’s underlying data infrastructure: He also optimized internal reports and data models for reliability, speed, and accuracy, ensuring the team could iterate quickly on insights even when application changes required lengthy security review.
• Modeled data for partner transparency: Lastly, the Fellow integrated his data pipeline and analysis improvements into the partner portal available to Login.gov’s agency customers. For this workstream, he spearheaded the delivery pattern for the data and worked across teams to iteratively incorporate features, and improve functionality and performance. The new reporting features in the portal enable agencies to access a wide variety of clean and interpretable metrics on relevant usage patterns and fraud outcomes and provide greater insight than was previously available to them.

The impact

The Fellow’s work has helped fundamentally change how Login.gov understands and improves its product. Reports that previously took hours to run and could fail during a run now complete in under 40 minutes with far greater reliability, thanks to refactored report delivery patterns. The shift to modeled, pipeline-driven data has made previously unreliable insights actionable and reproducible, directly informing which parts of Login.gov’s proofing processes need redesign (i.e. where data show that users are currently getting stuck), and has surfaced other opportunities now on the program’s roadmap.

This foundational work also lowers the barrier to data access across the Login.gov: product managers and other staff can now query modeled data directly, rather than needing deep knowledge of raw logs. In addition, the agencies now have self-service visibility into usage and fraud metrics.

The SSA field office pilot analysis also translated directly into action. Those findings informed three proposed changes to the agent-assisted proofing flow, demonstrating how data can shape proofing pathway design before a feature scales. Login.gov continues to coordinate these pilots with other federal agencies to test real-life applications of these updates.

The Fellow’s work also surfaced an important lesson for the Login.gov program more broadly : because application changes cascade into the data models powering product decisions, close coordination with across teams and shared definitions of concepts like “success” or “attempt” are as critical as the engineering itself.

Looking ahead, this reliable data foundation opens the door to predictive analytics that can flag fraud risk or personalize the proofing experience proactively, rather than just understanding behavior after the fact. Data continues to be embedded in every aspect of Login.gov, and so has the Data Science Fellow, who graduated from USDC into a full-time role on the Login.gov team.