By Danor Aharon — Senior Algorithms Developer
Will a borrower pay back its debt, is a fundamental question humans have been repeatedly asking themselves for about 5000 years since early societies adopted the concept of debt. Today, the global standard driving the lion’s share of debt in the global economy and public markets relies on rating agencies. The assessment of [a] corporate’s ability to serve debts is set to classes, known as risk ratings. Those risk ratings are the basis for the pricing of corporate debt by lenders.
The associated risk level, meaning the probability of default, is set by sequence of letters, where the highest yield rates, usually corresponding with default rates of > 1.5% are considered high yeild.
This standard approach is not available for the private tech companies, which are commonly known as startups. It’s important to point out that the focus of this article is on tech companies with pre-existing revenues. Some of what we call startup’s today, are actually medium size corporations that already generate hundreds of millions of dollars. Not really the classic definition or assumption of a startup.
While the total investments in later stage private tech companies are growing YoY, there’s still not a standard vehicle for assigning the risk associated with a private tech company defaulting on a debt facility. As a result, the debt facilities for private tech companies are lagging behind the market. Most tech companies are not using debt as a significant growth financing solution but are leveraging costly equity rounds to finance growth. Those are, as we are addressing in our article, actually have a very high cost of capital.
External Article (Raising Capital with Equity: How Expensive is Expensive)
The current risk assessment process is not set to enable scale. They are based on tribal knowledge and spreadsheet based reviews.
Private tech companies have several unique features that make the traditional corporate debt risk assessment models irrelevant. Many of the best private tech companies are choosing to run their business model with a negative EBITDA. By pushing the paddle on the growth engine, tech companies are knowingly creating negative cash flow in order to fuel exponential growth. Standard debt deployment models are not set to deal with companies supporting a growing negative cash flow. Most private tech companies are experiencing significant shifts in business models finding product market fit. They do not have years of stable operations to assess business resilience. It’s quite challenging to assess what would be the future revenue or cash balance of a private tech company, compared to an established business. Moreover, the assets and as a result securities are, in fact, quite soft. Most tech companies do not have large inventories or hard assets to leverage as securities. The ability to recover a debt using the [tech] IP, the most valuable assets tech companies possess, is severely limited. Measures of exposures, mostly loan to value ratios, can not serve as dominant decision drivers as they do in standard debt practice.
With an ever growing market opportunity and the large void in risk assessment capabilities for debt in the tech market, Liquidity is setting out to proliferate it’s risk assessment model to be the global standard for private tech debt risk ratings. We developed an ML model based on a proprietary scoring model. Using conditional probabilities, advanced statistical methods and sophisticated ML techniques which include at least 15 types of regression algorithms to multiple datasets, we are able to predict the probability of default andhe model is trained using data from over 20,000 companies from around the world, supplying a wide range of reference points for the model to learn from. Figure 1 shows one of the model scenarios where its algorithm examines and fits 10 different types of probability functions to a distribution of the fundability index of more than 15,000 companies (a fundability index is a normalized value between 0 and 1 that is developed by Liquidity’s Data Science group and represents the ability of a startup company to raise capital).
Figure 1: The distribution of the fundability index of 15,132 companies along with 10 types of probability functions fits.
The Liquidity model fuses real time data sets, integrated to a company’s core systems to drive it’s prediction. Using real-time data streams, reduces the variability and enables Liquidity to respond in real time to changes in risk profile or for a company’s financing needs. This novel approach to risk rating and opportunity scoring is probably applicable to other business verticals, and could revolutionize the $10T global corporate financing market.