🤖Webinar - From Theory to Practice: Real-World Applications of AI in Asset-Based Finance - Register here

Altin Kadareja



Private Debt data management process with CARDO AI

Private debt market actors cannot operate efficiently and stay ahead of competitors without the right data. Sourcing and assessing new investment opportunities, evaluating and managing risk involved with available investments, anticipating potential losses, and handling portfolio returns are all critical activities that require you to have consistent and comparable information.

At Cardo AI, our data scientists and engineers make sure that you always have the right data at your disposal, ready to be analyzed, and support you in taking timely and efficient investment decisions.

How do we do that? In this article, we are going to delve deep into our data management process and show you how we transform raw private debt data into actionable and valuable insights for our customers. 

Data management process at CARDO AI

At the moment, we are developing a standardized model for the private debt European market data, pulling real-time data points from 100+ different private debt data sources from 25+ distinct European countries.

Aggregating and handling different sources of data is the starting point of our data management process. Currently, our technology covers the entire cycle: data sourcing and treatment, data standardization, and data modeling. 

1. Data sourcing and treatment

Traditional sources of private debt data

Our engine collects and cleans private debt data from several sources as a first step.

Traditionally, actors in the private debt space use the following data types to come up with credit risk models (in a similar way as data used in the banking sector):

  • Company data: Country; Sector; Financial Ratios; Balance sheet data; Board and Executive Management data; Company’s organization data; Negative events; bankruptcy procedures etc.
  • Loan/Debt investment data: Investment and maturity date, amortization plan structure, interest rate; type of interest; expected return at maturity, etc.
  • Historical behavioral data: historical payment transactions; timestamp, amount, currency, type of the cashflow; missing payment data, etc.

Additional data

However, private debt investors have started to recognize the need for additional data, with two main objectives. First, protecting the downside with better credit risk models and real-time monitoring capabilities. And second, optimizing the upside with better pricing data in order to negotiate with confidence.

How can this be done? By using additional data. Here are some examples:

  • Standard borrower-related data: company identification numbers, company sector classifications, financial ratios, and balance sheet data from open banking transactions
  • Credit-related data, such as credit curve compositions for every borrower across credit originators, benchmarking pricing with companies in similar sectors, countries, or with comparable characteristics
  • Data from specialized third-party providers such as credit rating agencies, insurance companies, market data providers (European data warehouse), ESG rating providers, as well as alternative data sources (web footprint, product reviews, social data, etc.);
  • Loan and cash payment data from payment institutions and depository banks, as they can help us to understand the product characteristics and payment behaviors of the borrowers. 
  • External and alternative data, such as rating estimations, insurance protection calculations, and alternative data from our web footprint data extraction techniques (borrower size and real-time organizational information, product/service reviews, employee satisfaction level, social network presence, and news)

2. Private debt data standardization


As a second step, these data are transformed into a standardized format. This allows all users (from front-office, risk management, operations, and investment teams) to easily interact, use and compare information across standardized investment performance metrics. 

3. Private debt data modeling

As a third step, we model available data with different algorithms for our clients to trade and invest confidently. These proprietary machine learning models are the essence of our services because they can predict borrower behaviors not only with reference to the borrower but also to the credit product the investor is analyzing.

We have witnessed the explosive growth and novelty of credit products like Buy Now Pay Later, Revenue and Inventory financing, Salary Financing, and so on. All of these products cannot run on old metrics of the probability of default of the borrower over a one-year period of time. To properly assess the risk of these products, we need new logic and models. 

Our innovative models take into consideration the features and structure of the specific product under examination:

  • Loan characteristics, purpose, and maturity
  • The structure of the underlying purpose of the product’s usage
  • Channels that the loan has originated in

These and many other elements, combined, can generate better predictors for the borrowers’ behavior to pay back in time the loan. 

That is why at Cardo AI we have come up with ideas such as delay prediction models, propensity to pay back models, and revenue limit estimations. To discover our advanced models and AI-based products for faster scaling and better decisions, request a demo today.