Verified earnings for underwriting

The problem

Income that arrives from many platforms is invisible to a lender. A reliable earner looks unbankable because no single statement adds up to the whole picture.

Self-reported income is unverifiable, and pulling raw bank data is heavy, invasive and still not attested: you get transactions, not a signal you can stand behind.

And any score that touches a credit decision drags a vendor toward being a consumer reporting agency, with all the obligation that carries, a line most vendors blur and few state.

The solution

• Read a verified-earnings attestation: monthly buckets, a trailing-12-month verified total, stability and coverage. Only counterparty-verified income attests: self-reported and disputed amounts are reported separately and never blended in.
• Read the deterministic reliability score alongside it, with a factor breakdown. If your decision is a credit decision, /score/explain returns ranked, stable, versioned reasonCodes (the specific principal reasons an ECOA / Reg B adverse-action notice needs), and the catalog is public.
• Make the call on your side. Credda supplies evidence; the policy, the decision and the loss are yours. There is deliberately no decision, approve/deny or AI-verdict endpoint: that boundary is what keeps Credda outside FCRA consumer-reporting scope and keeps your risk model entirely yours.

How it works

The whole partner arc (connect a subject, report outcomes, subscribe to one webhook, read evidence, decide) is on the integrate page, with the FCRA boundary spelled out.

The API calls involved

• GET/api/v1/users/{id}/earnings/summaryVerified-earnings summary: trailing-12m verified total, median monthly, volatility, verified share.
• GET/api/v1/users/{id}/earningsFull attestation: monthly buckets, stability, coverage. Verified only; currency is null (platform-reported units).
• POST/api/v1/users/{id}/earnings/credentialA signed CreddaEarningsCredential the subject can present offline.
• GET/api/v1/users/{id}/score/explainFactor breakdown plus ranked reasonCodes: the Reg B / FCRA reasons behind the number.
• GET/api/v1/reason-codesPublic: the stable, versioned adverse-action reason-code catalog.
• GET/api/v1/scoring/modelPublic: the model itself, so you can audit any score against it. No key.

Credda attests a record; it never renders a lending decision. Where you take a credit decision, the reason codes are what let you meet your own ECOA / Reg B obligations.

# Verified earnings. Self-reported and disputed amounts are NOT blended in
$ curl https://api.credda.io/api/v1/users/subject_17/earnings/summary \
    -H "Authorization: Bearer $CREDDA_API_KEY"
{ "trailing12mVerifiedTotal": 48200, "medianMonthly": 3900,
  "verifiedShare": 0.82, "volatility": 0.19, "currency": null }

# The score, with ranked reason codes for an adverse-action notice
$ curl https://api.credda.io/api/v1/users/subject_17/score/explain \
    -H "Authorization: Bearer $CREDDA_API_KEY"
{ "factors": [ /* ... */ ], "reasonCodes": [ /* ranked, adverse first */ ] }

Full request and response shapes are in the API reference.

Start building