PD, LGD and EAD: expected loss explained for B2B credit teams

Banks have quantified credit risk with three variables for decades. The same three work just as well for a supplier granting open payment terms, and they turn an argument about whether a customer 'feels risky' into a number you can price.

Probability of default (PD)

PD is the chance that the counterparty fails to meet its obligation over a defined horizon, normally twelve months. It is what a credit score expresses: a score band maps to a default rate observed across companies with similar characteristics. PD is a statement about a population, not a prophecy about one company — which is why it is used to size exposure, not to predict individual behaviour.

Loss given default (LGD)

LGD is the share of the exposure you would not recover after a default, net of security, guarantees, retention of title and insurance. Unsecured trade creditors in an insolvency typically recover very little, so trade credit LGD is high unless mitigation is in place. LGD is the variable you control most directly: security, prepayment and credit insurance all reduce it without changing the customer.

Exposure at default (EAD)

EAD is the amount outstanding at the moment of default. For trade credit it is not simply the current balance: it is the balance you would have at the worst plausible moment — a full limit drawn, plus goods shipped but not yet invoiced, plus anything committed under an open order.

Putting them together

Expected loss = PD × LGD × EAD. On a €500,000 exposure with a 2% probability of default and a 70% loss given default, expected loss is €7,000 — a cost of doing business that should be priced into margin, not treated as a surprise. The formula's real value is comparative: it shows that halving the limit and taking security often reduces expected loss far more cheaply than refusing the customer.

Unexpected loss and concentration

Expected loss is the average. What damages a business is the tail: several counterparties failing together because they share a sector, a country or an end customer. That is why credit risk management pairs the expected-loss calculation with concentration limits — the arithmetic assumes independence that a real portfolio does not have.

Practical use for credit teams

You do not need a regulatory-grade model. Map your score bands to observed default rates from your own ledger, agree a default LGD assumption per mitigation type, and calculate EAD from limits rather than balances. Even a rough version makes exposures comparable, makes pricing defensible, and gives finance a loss provision grounded in something other than last year's figure.

— RisQo Research Team, Infocredit Group