Published in


Data-Driven Lending is Better

Written By: Darshan Vaidya (CEO of Credora)

Credora is a privacy-preserving credit oracle, designed to compute on the real-time risk of a borrower without needing to see the underlying data. The architecture produces cryptographic proofs that confirm both the computations being performed and the privacy of the data. Credora’s platform is designed to be a credit intelligence platform where the real-time data is contextualized with static financial information and helps lenders understand and validate a more holistic and up-to-date picture of a borrower.

Credora’s own credit team uses this approach to underwrite borrowers, and create dynamic credit scores that update immediately upon a change in creditworthiness. Those credit scores have outperformed other underwriting in the crypto space (with no defaults to date) and were able to accurately and immediately quantify the impact of the many shocks the crypto market suffered through 2022.

Over the course of the past year, many lenders we spoke to regrettably opted for a relationship-based lending approach as opposed to using real-time monitoring tools. They highlighted concerns around the completeness of data. As a business, it is important for us to address the misconception of having 100% visibility in credit, and emphasize the value of incremental data.

Relationship-Driven Lending

A common lender take: “Data on a borrower’s assets or liabilities in real-time is not useful, and in fact can be misleading if it does not capture 100% of the balance sheet. It’s more important to have a strong relationship with the borrower”

While a frequent conclusion, the logic is challenging to follow. Perhaps a valid starting point is to first look at the status quo. Lenders evaluate historical financial statements from a borrower, build a credit history, and validate some information through third-party attestations. The last part of that process (validation) has largely been ignored in crypto credit underwriting because, well, it’s notoriously hard to do. Audited financials are generally stale (and the whole crypto world has changed by the time the documents are ready), and real-time validation requires multiple integrations, getting borrowers comfortable sharing private data, and an understanding of how to make use of the data.

No one is suggesting that a lender throw away financial statements and rely solely on real-time data. Financial statements are a big piece of the puzzle when underwriting a business. But, the underwriting process should not take that document as truth, and instead validate it. When lenders refer to ‘100% coverage’, it is a flawed starting point. What exactly is 100% coverage using as the benchmark? The whole idea of financial statements and real-time data is to try to help construct a picture of what is likely to be 100% of the balance sheet, today. Lenders should avoid the perspective that balance sheets are the ‘100%’, dismissing incremental data if it does not fully prove that ‘100%’.

It is normal as an underwriter to approach data provided by a borrower with skepticism. If real-time data only captures 33% of the balance sheet assets, the conservative approach would be “the balance sheet is 3 times the size of the validated assets”, not “the real-time data only captures 33% of the balance sheet assets”. The approach involve skepticism towards the remaining 66%, considering those assets as very risky or non-existent, until proven otherwise.

In credit, especially in nascent sectors of credit, static data cannot be taken at face value. A complete approach requires piecing together multiple sources of information, being inquisitive and skeptical about that data (understanding that all data in a loan application is designed to look as favorable as possible), and being able to apply probability analysis on the likelihood of this data changing.

As an example of using incremental data, say you were to be betting on a sports event, and you could manage your exposure as the game unfolds. You could just lean on the score published every quarter, assuming you trust the publisher, and sit idly hoping the final score results in a favorable outcome. Or, you could watch the game on TV and adjust in real-time. That certainly seems like it would be useful. And let’s say the TV is broken — you can only see half the screen — do you switch it off, and assume it’s useless here? Or do you use what you see and hear, take it as a proxy of what is going on in the game, and manage your risk accordingly?

When a lender gets a financial statement at , and they can use Credora to validate 70% of it, it is then up to the lender to use other methods to validate the rest. Credora’s infrastructure can give them an good proxy of what the financial statement will look like in , or , and give the lenders the ability to react to market events. Certainly, this is a better proxy than the typical “hey just checking you aren’t affected by this UST thing?”. Manual diligence like this has its place, and is sensible, but it’s not even close to seeing real-time data that validates the manual assertions.

It’s debatable how much value 5% visibility gives vs 25% vs 75%. But, if seeing only 5% gives you discomfort as a lender, then you should simply demand more. If you ultimately fail to get more validation, then perhaps you should take as a data point, and maybe not lend?

Can partial visibility on a borrower, through Credora or otherwise, give a false sense of security? That can be said for any incremental piece of data, including balance sheets, bank statements, and screenshots of risk. If you approach the data assuming everything is perfect, any additional data point can provide comfort. It’s part of the underwriting procedure to approach all data with skepticism, and delve deeper into whether this data is representative of a borrower’s risk or not.

Credora’s novel technology gives you insight into data points that are verifiably accurate and untampered, and unlocks data that was previously considered too sensitive. Moreover, it is harder to manipulate this data as it is continuous, and not a snapshot in time. There was recent skepticism about proof of reserves for exchanges, given they have the ability to temporarily borrow assets to show liquidity, and then return them after the audit. This would easily be flagged when conducting real-time monitoring using Credora.

Credora’s data should be something every lender scrambles to add to their analysis, and request it from those not actively providing it.

It’s been odd feeling the need to write this explicitly, but given recent events, we can speak a bit more candidly. We strongly suggest lenders ask for data to validate the claims their borrowers are making. If borrowers won’t, or can’t provide it, lenders should read into that. If you do little to validate and your borrower is lying, don’t claim to be the victim on the back of a pinky swear.

Traditional Underwriting Process

One of the recent catchphrases we have heard is; “Maybe crypto isn’t ready for credit, traditional underwriting doesn’t work here”. Again, the broader issue was cherry-picking traditional underwriting methods and then scaling it while there were still substantial holes in it. It’s hard to say something doesn’t work when you didn’t actually do it properly.

Here is a broad list of the normal steps when conducting due diligence on a counterparty:

  • KYC & Entity Due Diligence
  • Historical Financial Statement Analysis
  • Business Model Overview
  • Revenue Analysis
  • Liabilities evaluation alongside interest coverage
  • Third-party validation of the above
  • Monitoring of changes to the above

If you would like to read the traditional guidelines around underwriting, the European Banking Association (along with the OECD and the OCC) have written their opinions on the standards to be upheld.

Standard credit diligence requires

It is fair to say that we haven’t really given ‘traditional underwriting’ a fair crack, and we strongly believe it’s worth giving it an actual effort before turning our noses up at the rest of the multi-trillion dollar credit industry.

The prevailing lower standards and relationship-based lending transpired because the data being handed to lenders wasn’t perfect, and incentives in venture-funded lending startups mean that it is more convenient to ignore that data. These lenders make money by offering retail and institutional investors a yield, and then absorb liabilities across multiple borrowers aiming to capture a spread. This creates direct balance sheet risk for them, and the ultimate lenders have no transparency regarding the strategies deployed by principal lenders to generate yield, or the associated risk. These dynamics together incentivize maximizing the spread to drive profitability, and the opacity limits the incentive to reduce the risk.

“Well, crypto is a nascent space, and there is limited data available to make these assessments, especially as counterparties have limited track records etc.”

This, I believe, underlines the frustration we (as a Credit Intelligence platform) feel, having tried to give this data to lenders for basically free for over a year. There incremental data available on borrowers, and that data can be obtained for those looking for it.

Credora provides unparalleled access to real-time data on a borrower’s on-chain and off-chain assets, and the ability to validate the balance sheets cheaply and continuously. If there are gaps in the data, there should be an urgency on the lender’s side to force borrowers to share more data to plug those gaps.

Moreover, there is a swath of data available on-chain relatively cheaply (Nansen, Chainalysis, TRM Labs), and if cost is a concern, the thrifty amongst us can find the data for free (block explorers) that provide many insights on certain borrowers’ activities. If analyzed correctly, this data could have given insights into the risks borrowers were taking on-chain, and moreover, the ability to notice transactions between them and known lender wallets, giving some insight into their liabilities.

There may be a reason why borrowers can’t validate their balance sheet in real-time. For example, are many of the positions OTC? That is certainly a possibility, but also relatively easy to provide alternative evidence to support the claim. Are the positions mostly illiquid e.g. ventures? That is also certainly a possibility, but it should almost certainly change your lending approach. Or is it some other reason that revolves around not having time to validate positions on Credora and the dog ate my list of API keys? (While I say this in jest, some of the excuses from certain now bankrupt borrowers will live in a private hall of fame).

It is up to us as an industry to enforce higher standards, and set a high enough threshold in terms of transparency. Borrowers share the information if they to share it to get the loan. It’s up to lenders to set that ‘need’.

So What Next?

Nobody sets out to lose money, and generally, I have to assume lenders are well-intentioned. However, the structural issues with principal lending models were discussed earlier — risk opacity and AUM incentives lead to underwriters shooting for unsustainably high yields and taking incrementally higher risk. At Credora we think that model of lending will be a thing of the past.

Credora was founded to disintermediate credit markets by aggregating unique data on borrowers while respecting their privacy. We take the opposite approach of the principal lender model. It focuses on transparency and optionality for the ultimate lender. Ultimate lenders get access to data on all their ultimate borrowers, and access to the infrastructure needed to seamlessly lend to whichever borrower they choose. We have facilitated $1bn in undercollateralized loans since inception, and have been able to successfully navigate one of the most challenging credit markets.


This is quite simply an argument for data-driven lending and a call for higher underwriting standards. The issue hasn’t been poor or incomplete data, it’s an unwillingness to ask or look for it, and an inability to use it when it’s there, and instead deference to relationships and a blind pursuit for market share. We think there is a better way, with institutional-friendly, licensed rails that offer thorough credit evaluation, transparency, and real-time monitoring of borrowers as standard.



The Private Credit Oracle. Credora is an end-to-end lending solution facilitating credit by validating real-time risk metrics in a zero-knowledge environment.

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store

The Private Credit Oracle. Credora is an end-to-end lending solution facilitating credit by validating real-time risk metrics in a zero-knowledge environment.