By Darshan V. —
Onchain asset management has built out its asset side. The risk infrastructure around those assets is considerably less developed.
Onchain asset management has built out its asset side. Stablecoin supply has roughly doubled to $300 billion, tokenized real-world assets have crossed $30 billion, and BlackRock, NYSE, DTCC, and BNY Mellon have each shipped meaningful products on public blockchains over the last eighteen months. The risk infrastructure around those assets is considerably less developed.
How risk works in traditional fixed income ==========================================
It is worth being concrete about what institutions are used to before looking at what is available onchain.
A corporate bond issued tomorrow goes through a standard process before an allocator sees it. Moody's, S&P, and Fitch each assign a letter rating from AAA down to D, encoding a probability of default over a defined horizon and calibrated against decades of realized defaults on comparable issuers. The methodology is public, subject to regulatory oversight, and material changes have to be disclosed.
Alongside the rating, the sell-side publishes a credit spread: the yield the bond pays above a risk-free benchmark, compensating the holder for expected loss plus a liquidity and uncertainty premium. A wider spread reflects more perceived risk. The spread is observable, comparable across issuers, and updates in real time.
Inside an allocator's portfolio, each position carries a value at risk, typically the 95th or 99th percentile loss over a defined period, computed from historical volatility and correlations with everything else in the book. VaR is how a portfolio manager answers the question every risk committee asks: if next month is a bad month, how bad can it get.
These terms were standardized over decades through the interaction of regulators, rating agencies, central banks, and sell-side research. When a PM at Fidelity compares a new high-yield issue to an existing position at BlackRock, they are reading the same ratings and spreads, and calculating VaR the same way. That shared vocabulary is what makes institutional allocation possible at scale.
The onchain equivalents are at an early stage of development.
What institutional allocators run into ======================================
An allocator running diligence on an onchain strategy will find that a handful of agencies have begun rating strategies, that observable and comparable credit spreads do not yet exist, and that VaR is calculated differently on every platform. Most platforms publish their own metrics, on their own scales, with assumptions that are rarely documented in full.
In practice, this surfaces as five recurring diligence blockers.
Rating scales are not comparable across platforms. A BB-rated corporate bond carries a specific meaning calibrated against historical default data. A "Conservative Onchain Strategy" label does not. Some platforms publish a probability of loss; others publish only an APY target. Benchmarking across platforms is difficult in the absence of a shared benchmark.
Leverage is typically expressed as a multiplier rather than as expected loss. A "3x leveraged" label describes capital efficiency. Expected loss under leverage depends on the liquidation buffer and the depth of secondary liquidity available during unwind. A 3x position in a market with a wide LLTV buffer and deep exit liquidity behaves very differently from a 3x position in a thin market with a hardcoded oracle.
Oracle risk is scored on a single axis. Most frameworks treat oracle choice as a question about price feed accuracy. The question that drives expected loss under leverage is whether the oracle helps or hurts the holder given which side of the trade they are on. A hardcoded peg reduces liquidation risk for a borrower; the same peg traps a lender in an undercollateralized position during stress.
Tokenized real-world assets are modeled as their off-chain equivalents. A tokenized Treasury bill is sovereign credit, redeeming T+1 against a NAV oracle that updates on a schedule. Between updates, the protocol sees one price while the real-world market may have moved. When the same position is posted as leverage collateral, that gap becomes a liquidation blind spot with no off-chain equivalent. Most frameworks model the PD and miss the interaction.
Risk models themselves are not governed. Traditional credit agencies publish methodologies, back-test them, and have to disclose material changes. Onchain risk outputs are often opaque, sometimes by design and sometimes because the methodology page has not been updated in eighteen months. There is no regulatory framework behind an onchain rating comparable to the one behind a Moody's rating.
Any one of these is enough to stall diligence. Together, they explain why decisions that should take four weeks take six months, and why the volume of "DeFi risk management" PDFs continues to grow without closing the underlying gap.
What a credible onchain risk layer would look like ==================================================
The work required is largely a faithful translation of the traditional risk toolkit.
A credible onchain risk layer would publish unified methodologies across every strategy it rates, expressed in a single vocabulary on a single scale. Risk would decompose to the position level, so an allocator can see which positions drive a vault's expected loss rather than reading only a top-line number. Oracle choice would be evaluated from both sides of the trade, and be anchored on what matters; how the choice of oracle affects the likelihood of the investor losing money. Leverage, oracle update frequency, and redemption windows would be modeled together, since they interact under stress. The model itself would be governed: versioned, and when sufficient data exists, back-tested against realized loss events, and independently audited.
Where Railnet fits ==================
Railnet is the operating layer for onchain asset management. The risk data room is the component of that layer directly addressed to the gaps described above.
While a ratings product is challenging to meaningfully calibrate given data around DeFi losses is nascent, and the losses to date are more binary in nature, we believe that granular transparency and tools to quantify risk is vital. Every strategy built on Railnet is analyzed using a universal approach to risk, and focuses on surfacing the various risk factors in DeFi in a unified manner. Risk decomposes in three stages, covering asset credit, strategy risk, and vault aggregation, with outputs expressed as probability of any loss of principal, or maximum loss at the 95th percentile. On top of this, while technical risk is binary in nature, we believe surfacing visual representations of the risk vectors can ensure users have a full understanding of the risk vectors they are exposed to when deploying in any vault.
Railnet's risk framework and emphasis on transparency is the foundation to scalable vaults. The work that follows includes published stress scenarios against realized events, post-mortems on positions that experience close calls, and queriable data rooms that make analysis more simple for an asset manager or depositor. Ultimately, these building blocks are what is needed to meet the standard institutions are used to.
If you are running diligence on an onchain strategy and would like to see a position walked through the methodology, the full methodology and a per-strategy decomposition pack are available on request.