Built by practitioners, for practitioners
BankAtlas began with a question: could bank risk be measured earlier and more systematically from the underlying regulatory data?
It started before 2008. In the years leading up to the financial crisis, our founder — a veteran investor and portfolio manager — was analyzing U.S. banks and saw risks that conventional measures did not always fully capture. The crisis reinforced the need for a more granular, quantitative approach. The work that followed focused on building a bank-level historical dataset and analytical framework, quarter by quarter, designed specifically to measure those risks.
Over nearly two decades, that dataset grew into a quantitative engine for gauging the risk of every U.S. bank — individually and across the system — through proprietary grades and scores. It was refined across real market cycles, from 2008 to the regional-bank failures of 2023, and built to answer the questions that actually matter about a bank’s profile.
2023 was the test. In point-in-time historical testing, the current models identify the funding and fundamental weaknesses visible before the 2023 failures. First Republic had carried a bottom-tier composite grade since 2021 — more than two years before it failed — and its run-vulnerability climbed to the top of its peer group by early 2023. Silicon Valley Bank, whose composite looked healthy to the end, is identified by the second lens: a run-vulnerability signal pointing to the funding fragility the headline grade missed. The current model specification is the one that runs on every bank today.
BankAtlas is a quantitative bank-intelligence platform. Our grades and scores are proprietary analytical measures — not credit ratings — and we don’t give investment advice; we provide general, data-driven bank intelligence, now enhanced with AI that makes our proprietary grades queryable and interpretable. Built by practitioners, for the analysts, investors, and bankers who ask the same questions we do.