The LendingRobot Score

Why was LendingRobot Score developed and how does it work?

LendingRobot Score is our own mechanism for ranking loans against each other. We developed it because we believe that different borrowers, regardless of the grade or sub-grade that a platform might assign to the borrower’s loan, have different risks of defaulting. LendingRobot Score uses statistics, survival analysis, and machine learning to output a score which can be used to select what we think is a better loan.

Why use LendingRobot Score?

The LendingRobot Score looks at borrower characteristics (FICO, debt-to-income, credit lines open, etc.) to estimate the probability of if and when said borrower would default. This gives investors a method of filtering loans that looks at all borrower characteristics simultaneously. LendingRobot Score can be used alone or in combination with other filters to increase selectivity and potentially increase returns.

Why is LendingRobot Score different from Expected Return?

LendingRobot Score essentially differs from Expected Return through the probability of default. Expected Return uses historical probabilities of default from the platform based on sub-grade, whereas LendingRobot Score uses the historical probability of default and adjusts it up or down based on borrower characteristics. Expected Return is a platform-based estimate of return while LendingRobot Score is LendingRobot’s measure of loan rank.

Some of the language in the old blog posts may be confusing, as “LendingRobot Score” used to be referred to as “Expected Return” until mid 2015. The downside to using it like this is that the score was never a neutral measurement for our clients, as the same algorithm was used for both selection and reporting. Hence we changed ‘expected return’ to the neutral measurement that is presented today, LendingRobot Score’s role shifted entirely to selecting better loans. So the LendingRobot Score should not be confused with an expected return.