AnaCredit for payment firms: scope, thresholds and the 95-attribute data model

“AnaCredit is bank-only” is one of the most expensive misconceptions in EU regulatory operations. The European Central Bank’s analytical credit dataset captures granular credit-exposure data from across the euro-area financial system — and payment institutions, e-money institutions and crypto-asset service providers that hold credit exposures can fall in scope quietly, depending on how their national central bank classifies them. The reporting threshold is per debtor, not per reporter; the data model runs to 95 attributes across ten datasets; and the data-quality bar is among the highest of any regular ECB return. This piece maps the scope question, the thresholds, the data model, and two worked examples showing where non-bank lenders land.

1. What AnaCredit is

AnaCredit is the analytical credit-and-credit-risk dataset operated by the European Central Bank, established by Regulation (EU) 2016/867. It collects granular, loan-by-loan information on credit exposures held by reporting agents in the euro area. National central banks — Banco de España, De Nederlandsche Bank, Banque de France, Banca d’Italia, Deutsche Bundesbank, Banque centrale du Luxembourg and their peers — operate as the local reporting interface: data flows from the reporter to the national central bank, then up to the ECB’s central database.

The dataset underpins ECB monetary-policy analysis, financial-stability monitoring and supervisory work under the Single Supervisory Mechanism. It also feeds national credit registers — CIRBE in Spain, the Banque de France’s central risk service, Banca d’Italia’s Centrale dei Rischi — see the CIRBE piece for the Spanish national overlay.

2. Why “bank-only” is wrong

Article 3 of the AnaCredit Regulation defines reporting agents around credit institutions resident in a reporting Member State. That is the headline, and the source of the confusion. Two qualifications matter:

• Foreign branches of euro-area credit institutions are also reporting agents, even for activity booked outside the euro area.
• The classification is statistical, not licence-label-based. Entities classified as deposit-taking corporations under the relevant ECB statistical framework can be caught, and national central banks differ in how they read the perimeter for non-bank lenders.

For most pure payment-services EMIs and PIs, AnaCredit is genuinely out of scope. The moment the entity holds a credit exposure — overdraft, credit line, buy-now-pay-later receivable, factoring asset, repo — the analysis tightens, and the answer becomes jurisdiction-specific.

3. The reporting thresholds

Two thresholds determine what gets captured:

• €25,000 per debtor — the headline AnaCredit threshold. A debtor whose aggregate credit exposure to the reporting agent reaches this level in the reference period is in scope.
• Instrument-level granularity — once the debtor crosses the threshold, the reporting is loan-by-loan across that debtor’s instrument set, down to small individual instruments.

The threshold is computed per debtor across all of that debtor’s instruments. A retail BNPL portfolio with balances mostly under €25,000 may legitimately have very few in-scope debtors; a corporate-lending or factoring book of any size will be substantially in scope.

4. Worked example — an EMI offering business overdrafts

Facts: a euro-area e-money institution serving small businesses adds an overdraft feature. Most limits are small, but a few hundred customers run aggregate drawn exposures above €25,000. The firm has never filed a statistical credit return.

What the rule says: the overdrafts are credit exposures, and the debtors above €25,000 would be reportable loan-by-loan — if the EMI is a reporting agent. That is the pivotal question, and it is answered by the national central bank’s classification of the entity under the ECB statistical framework, not by the firm’s own reading of “credit institution”.

What the practitioner does: puts the scope question to the national central bank in writing before the product scales, with the product structure and projected exposures laid out. If the answer is in scope, the firm plans a real implementation project — counterparty reference data, LEIs, the attribute model — rather than a spreadsheet. If out of scope, the written answer goes in the file, because the classification can be revisited as the lending book grows.

5. Worked example — a BNPL portfolio of small balances

Facts: a payment institution operates a consumer BNPL product. Average outstanding balance is a few hundred euro; no single consumer’s aggregate exposure approaches €25,000. Total portfolio, however, is large.

What comes back: the €25,000 threshold is per debtor, not per portfolio. However large the book, if no individual debtor’s aggregate exposure reaches the threshold, there is nothing to report at instrument level under AnaCredit — even for an entity that is a reporting agent. National credit-register rules may still capture the exposures separately under their own, sometimes lower, national thresholds.

What the practitioner does: builds a monthly per-debtor aggregation check anyway. The exposure calculation must aggregate across all instruments of the same debtor, so a consumer who stacks several purchases plus an overdraft could cross the line. The check is cheap; discovering in an inspection that reportable debtors existed for a year is not.

6. The 95-attribute file

For each in-scope instrument, AnaCredit captures up to 95 attributes across ten datasets:

• Counterparty reference data — debtor identifiers, sector classification, country of residence
• Counterparty risk data — probability-of-default class
• Counterparty default data — default status and date
• Instrument data — type, currency, maturity, original and outstanding amounts
• Financial data — accrued interest, accounting classification
• Joint-liability data — multiple debtors on the same instrument
• Accounting data — provisions, impairment staging
• Protection received data — collateral type and value
• Instrument-protection received data — links between instruments and protections
• Counterparty-instrument data — the debtor’s role on the instrument

The attribute list sits in the annexes to the Regulation, each attribute with defined permissible values. Validation at the national-central-bank ingest is strict, and the data layer is the long pole of any first implementation — see the AnaCredit data model for the attribute-level walkthrough.

7. Cadence, channel and data quality

• Frequency: monthly, with the reference date at the last calendar day of the month; some datasets update less frequently.
• Deadline: set by national rules — check your national central bank’s reporting calendar.
• Channel: via the national central bank’s reporting infrastructure — in Spain alongside the CIRBE flows, in the Netherlands through DNB’s digital reporting portal, in Germany through the Bundesbank’s statistical reporting framework.
• Data quality: quality metrics are tracked per reporting agent, and persistent data-quality issues escalate into a supervisory matter, not just a technical one.

8. The LEI requirement

Legal-entity debtors reported in AnaCredit must be identified consistently, with the Legal Entity Identifier as the anchor identifier for entities that have one. For large corporates this is rarely an issue; for smaller commercial debtors and special-purpose entities the identifier may not exist at first reporting, and the reporter needs a process to procure or trigger assignment. This is the same LEI discipline that the DORA Register of Information and Article 30 impose on ICT third-party arrangements — firms in scope of both benefit from a single internal LEI register.

9. Where the national differences live

AnaCredit is one ECB Regulation but many national reporting flavours:

See NCBs compared for the detailed matrix.

FAQ

What to do, today

• Head of product: map the product set against the AnaCredit notion of credit exposure — overdrafts, credit lines, BNPL, factoring, repos.
• Head of reporting: if any are present, put the reporting-agent question to your national central bank in writing and file the answer.
• Data engineering: scope the attribute model early — the counterparty and instrument data layer is the long pole of a first submission.
• Operations: procure LEIs for corporate debtors that lack one and bake LEI capture into onboarding.
• Architecture: feed the national credit register and AnaCredit from the same source-of-truth dataset — duplicating the flow is wasted engineering.

Related: AnaCredit phases · The AnaCredit data model · Forbearance reporting · NCBs compared · COREP under IFR/IFD · What is CIRBE? · DORA Register of Information · Where to base your EMI · DAC8 XML reporting — the data build