How likely is it to be repaid?

KivaScore uses a machine-learning model trained on historical Kiva loans to estimate how likely a new loan is to be repaid. A higher score means the model sees a lower chance of default.

The estimate is based on information such as:

• The loan itself: amount, repayment length and schedule, and what the money will be used for
• The borrower or group: group size and gender composition
• The Kiva field partner: its past default rate, lending history, typical loan size, and funding speed
• The surrounding environment: the country's economy and governance, plus the borrower's location and climate conditions

Repayment matters because repaid funds can be recycled into new loans, allowing the same capital to support more borrowers over time.

The explanation shown after scoring describes what pushed the estimate up or down; it does not change the score.

How much impact could it have?

KivaScore looks at four different ways a loan could matter, scores each one separately, and then combines them into one impact measure.

The first three scores come from structured rubrics applied to the loan listing: what the borrower is financing, the type of activity, and the description. Vulnerability combines explicit signs of household hardship with World Bank regional poverty data matched to the borrower's location.

The final 0–100 number is a percentile. An Impact Score of 80 means the loan ranks around the 80th percentile compared with recent Kiva loans.