zLend: A Dual-Scope Cash-Flow Reconstruction Framework for On-Chain Credit Underwriting
This paper introduces zLend, a deployed on-chain credit underwriting framework that reconstructs wallet cash-flow histories through dual-scope analysis of stablecoin and total fungible transfers to generate distinct liquidity and repayment-capacity signals, enabling risk assessment without traditional income verification.
Original paper dedicated to the public domain under CC0 1.0 (http://creativecommons.org/publicdomain/zero/1.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
In the world of digital finance, money moves through public ledgers that record every transfer, yet these records often tell only a partial story. Traditional banks rely on credit bureaus to judge a borrower's ability to repay, using verified income, past debt history, and a score that aggregates behavior across many lenders. In the decentralized world of digital currency, no such bureau exists. A borrower has no credit file, no verified paycheck, and no record of past loans. All that is available is a permanent, public log of transactions. For years, the only way to judge a borrower in this space was to look at their total wealth: the sum of every digital asset they held. This approach, however, suffers from a fundamental flaw. It treats all money as equal, ignoring the difference between cash that is ready to spend and assets that are locked up, volatile, or difficult to sell. A person holding a large amount of a speculative token might appear wealthy on paper, but if that token crashes in value or cannot be sold quickly, they may have no actual ability to repay a small loan due next week.
A team of researchers has developed a new system called zLend to solve this problem by looking at the flow of money rather than just the total amount. Instead of simply adding up a wallet's contents, the system reconstructs a daily history of what that wallet could actually spend. It does this by tracking every incoming and outgoing transfer to build a timeline of the wallet's balance, day by day. The researchers realized that to understand a borrower's true capacity to repay, they needed to look at the wallet through two different lenses simultaneously. The first lens focuses only on stablecoins, which are digital currencies designed to hold a steady value, like a dollar. The second lens looks at the wallet's entire portfolio, including all other tokens and assets. By comparing these two views, the system can spot dangerous mismatches. If a wallet shows a massive total value but almost no stablecoins, the system recognizes that the borrower is likely unable to repay a loan in cash, regardless of how rich they appear on paper.
The researchers built this framework to operate entirely on public data, without needing any outside information or human approval. They created a method to take a raw list of transfers and turn it into a clean, continuous record of daily balances. This reconstruction starts from scratch, inferring the wallet's opening balance based on the first few transactions and ensuring the balance never drops below zero. From this reconstructed history, the system calculates several key signals. It measures how often the wallet had enough stablecoins to cover a specific loan amount, how steady the income flows are, and how much the balance dropped during its worst periods. It also looks for patterns that resemble a salary, such as regular payments arriving from the same source at consistent intervals. These signals are then combined to assign a borrower to a specific tier, ranging from strong to weak, which determines their eligibility for a loan.
The most critical finding of the study is that looking at total wealth alone is a poor predictor of repayment ability. The researchers tested their system against a set of carefully constructed examples, including a wallet that held a large sum of volatile assets but very little cash. Under a traditional system that only checks total value, this wallet would look like an excellent borrower. Under the new zLend system, it was correctly flagged as a high risk because its liquid reserves were insufficient to cover even a small loan. The system successfully identified that the borrower's wealth was not accessible for repayment. Conversely, the system also recognized wallets with modest total wealth but a steady, reliable flow of stablecoins as strong candidates. This dual-view approach allows the system to distinguish between someone who is truly solvent and someone who is merely holding assets that cannot be used to pay a bill.
The researchers also discovered that the decision to grant a loan depends heavily on the size of the loan being requested. A wallet might be considered a strong borrower for a small loan of ten dollars but a weak borrower for a loan of ten thousand dollars, simply because its cash reserves are not large enough to cover the bigger amount. This sensitivity is not a flaw but a feature, reflecting the reality that a borrower's ability to repay is relative to the debt they are taking on. The study further showed that different risk factors operate independently. A borrower might have a perfect record of having enough cash on hand but still be rejected because their balance recently crashed by a large percentage, indicating instability. Another borrower might have a stable balance but fail because they have no history of regular income. The system treats these as separate, non-interchangeable criteria, ensuring that no single weakness is overlooked.
To ensure the system works correctly in the real world, the researchers built it with extreme precision. They implemented the logic in two different computer languages and verified that both versions produced identical results down to the smallest decimal place. This rigorous testing confirmed that the mathematical rules for calculating balance, volatility, and risk were applied consistently. The system is already live, integrated into third-party lending applications where it makes real-time decisions on actual loans. By turning a chaotic stream of public transactions into a clear picture of daily liquidity, the framework provides a way to extend credit without collateral, relying instead on the behavioral evidence of how a borrower manages their money day to day. The work demonstrates that with the right reconstruction of history, it is possible to infer creditworthiness from public data alone, offering a path toward a more inclusive and data-driven financial system.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.