Data sources

CredAI scores whatever implements the data contract. Today that's the built-in generator; pointing it at a real API is configuration, not code.

API reference

Active provider

Synthetic

Checking…

Switch to real data

1. Expose GET /customers, GET /customers/:id and optionally GET /applications returning the contract below.
2. Set three environment variables and redeploy.

.env.local

CREDAI_DATA_PROVIDER=http
CREDAI_DATA_API_URL=https://data.your-bank.example/v1
CREDAI_DATA_API_KEY=••••••••   # sent as a Bearer token

Try it locally: npm run mock:api starts a partner API with a different population on port 4010.

Data contract

Validated with a Zod schema on every record. Invalid records are rejected and counted, never coerced.

Utility bills
utility.bills[]: issuedAt, dueDate, paidAt | null, amount, kind, provider
LESCO, K-Electric, SNGPL
Mobile account
telecom.accountOpenedAt, payments[]: date, amount, kind, status
Jazz, Zong, Telenor, Ufone
Mobile wallet
wallet.transactions[]: date, amount, direction, status, balanceAfter
JazzCash, Easypaisa, SadaPay
Bank account
bank.transactions[]: date, amount, direction, category, balanceAfter
Any account with statement access
Repayment history
credit.installments[]: dueDate, paidAt | null, amount
Microfinance banks, BNPL

Feature catalogue

17 features engineered from the contract over a trailing 12-month window. The same code runs in training and in production.

Feature	Source	Definition	Weight
Utility bills paid on time `utility_on_time_ratio`	Utility bills	Share of utility bills paid on or before the due date.	+0.585
Utility bills left unpaid `utility_unpaid_ratio`	Utility bills	Share of bills still unpaid more than 30 days after the due date.	-0.062
Utility history length `utility_months_observed`	Utility bills	Months in the 12-month window with at least one bill.	0.000
Mobile account activity `telecom_active_ratio`	Mobile account	Share of months with at least one successful top-up or bill payment.	+0.174
Mobile number age `telecom_tenure_months`	Mobile account	Months since the mobile account was opened.	+0.345
Failed mobile payments `telecom_failed_rate`	Mobile account	Failed top-ups or bill payments as a share of all attempts.	-0.068
Wallet income regularity `wallet_active_ratio`	Mobile wallet	Share of months with incoming wallet transfers.	0.000
Failed wallet payments `wallet_failed_rate`	Mobile wallet	Failed wallet transactions, usually from insufficient balance.	0.000
Wallet balance cushion `wallet_buffer_ratio`	Mobile wallet	Average month-end wallet balance relative to monthly wallet spending.	+0.125
Income regularity `income_regularity`	Bank account	1 − coefficient of variation of monthly account inflows.	0.000
Income covers spending `cash_flow_coverage`	Bank account	Total inflows divided by total outflows over the window.	+0.134
Spending volatility `expense_volatility`	Bank account	Coefficient of variation of monthly outflows.	-0.222
Overdrawn days `negative_balance_days`	Bank account	Average days per month the account closed below zero.	-0.208
Savings buffer `buffer_months`	Bank account	Average month-end balance divided by average monthly spending.	+0.197
Instalments paid on time `repayment_on_time_ratio`	Repayment history	Share of previous loan instalments paid by the due date.	+0.203
Days late on instalments `repayment_avg_days_late`	Repayment history	Average days late across previous instalments (0 when on time).	-0.074
Data sources connected `data_coverage`	Derived	Share of the five supported sources the customer has connected.	+0.167