Skip to main content

Risk Score

Risk Score provides a risk signal derived from customer transaction and account data. It is designed to support underwriting decisions by offering a consistent, explainable assessment of a customer’s probability of default over time.

Retrieving this report

Risk Score is returned as the risk entry of the report bundle - one call returns it alongside Completeness and Aggregate. Check the entry's status before reading its data; the fields below describe that data payload.

warning

Read Completeness from the same bundle alongside Risk Score. Insufficient or gappy transaction data can reduce model accuracy and lower confidence in the output.

Interpreting results for credit

  • Treat value and rating as one signal in a broader decision framework - always combine with Completeness and Aggregate from the same bundle
  • Use shapleyValues to understand which features drove the score, supporting audit trails and decision explainability
  • A null value or populated errorCodes indicates the model did not have sufficient data - escalate these cases for manual assessment
  • Review pdConfidenceRating alongside rating - a low confidence score means the result should be weighted accordingly

Response fields

Score and rating

FieldDescription
valueNumeric risk score (probability of default). May be null if insufficient data is available
ratingRisk band derived from the score, on a credit-style scale from strongest to weakest: BB+, BB, BB-, B+, B, B-, CCC, CC, D. null when no score could be produced
pdConfidenceScoreProbability of default confidence score
pdConfidenceRatingNamed confidence band for the PD score

Model features

The features object contains the input signals used by the risk model:

FieldDescription
arbAverage running balance
averageSalesAverage sales or income amounts
dishonourCountNumber of dishonoured transactions in the assessment period
averageTaxPaymentAverage tax payment amounts
averageFinancialCountAverage count of financial transactions

Explainability

FieldDescription
shapleyValuesMap of feature names to their model contribution values. Use these to explain the key drivers of the risk score

Transactions metadata

The transactionsMetadata object describes the data the score was calculated from:

FieldDescription
earliestTransactionDateEarliest transaction date used in the assessment
latestTransactionDateLatest transaction date used in the assessment
transactionCountNumber of transactions used in the assessment

Time series

The timeSeries object provides a temporal breakdown of the risk signal:

FieldDescription
startDateStart of the time series window
endDateEnd of the time series window
frequencyTime interval for each data point
windowDaysSize of the rolling window, in days
valuesArray of points, each with date, pd, rating, features, and warnings

Error information

FieldDescription
errorCodesList of codes returned when the model cannot produce a score - for example, due to insufficient transaction volume

Next steps

  • Reports - How to retrieve this report and read its status
  • Aggregate - Complement risk assessment with cashflow and affordability analysis
  • Completeness - Review data coverage if the score is null or confidence is low
  • API Reference - Full endpoint schema and parameter details