Analysis
The analysis relies on data gathered by the underlying categorization process, so the categories that appear in the analysis response are the E1 categories. The full list per profile is documented on the E1 categorization categories page, and it is identical for every supported countryCode.
This page describes how those categories, confidence scores and flags are grouped into the analysis sections.
Client profile
clientProfile groups transactions into buckets that characterise the customer. The buckets differ per profile, and every bucket is always returned — a bucket with transactionCount 0 means nothing matched.
Retail buckets
internal.othertax.vehicle, expense.transport.vehicle, insurance.vehicle, expense.transport.fuel, expense.transport.parkingexpense.essential.childcare, subsidy.familyalimony.income.pension., plus investment.pensionsubsidy.unemploymentloan., except loan.debt_collectionloan.debt_collectionincome.salaryhousing.rent, housing.mortgagesubsidy., except subsidy.unemployment, subsidy.family and subsidy.family.incomeother.finesgambling.otherinvestment.cryptoCorporate buckets
tax.loan., except loan.debt_collection, loan.vehicle_lease and loan.equipment_leaseloan.debt_collectionloan.vehicle_leaseloan.equipment_leaseinternal.othergambling.otherinvestment.cryptoFlagged transactions
flaggedTransactions groups transactions by the flags the model assigned. The bucket name is the readable label; the flag is the value that appears in the flags array of the transaction level output.
reversal_paymentrefundpayment_arrangementLow confidence transactions
lowConfidenceTransactions groups transactions by how certain the model was about the category it assigned, using the categoryConfidence score. Use these bands to decide which categorizations are worth reviewing.
0.500.50 up to but not including 0.700.70 up to but not including 0.80Transactions categorized with a confidence of 0.80 or higher do not appear in any of these buckets. The bands cover low confidence results only, so they are not a complete partition of the analysed transactions.
Other highlighted transactions
otherHighlightedTransactions singles out transactions using heuristics rather than category or flag.
10000, with no decimal part, and an exact multiple of 10000. The sign is ignored, so both inflow and outflow qualify.Credit quality metrics
creditQualityMetrics is reserved for future use. E1 currently always returns an empty array. The DriverEntry shape is documented so integrations can be written against it ahead of time.