The AnaCredit data model: ten datasets and ~95 attributes explained
The AnaCredit data model is where implementation projects live or die. Roughly ninety-five attributes, organised into ten linked datasets, reported monthly at instrument level — with referential-integrity rules that reject the whole record when one identifier breaks. This piece maps the ten datasets in plain language, flags the attributes that cause the most rejections, and explains the identifier discipline (LEI, national identifiers, protection IDs) that has to be in place before the first file is generated.
1. The architecture in one view
AnaCredit reporting under Regulation (EU) 2016/867 is built on two templates and ten datasets. Template 1 carries the counterparty and instrument core; Template 2 carries the accounting and risk view. Every record links back to three master identifiers: the reporting agent, the counterparty and the instrument (contract ID + instrument ID).
2. The ten datasets, in plain language
| # | Dataset | What it carries |
|---|---|---|
| 1 | Counterparty reference | Who the debtor (and protection provider, servicer, originator) is — name, identifiers, legal form, sector, country, size class |
| 2 | Instrument | The contract — type, currency, inception date, maturity, purpose, interest-rate basis |
| 3 | Financial | The monthly numbers — outstanding nominal, accrued interest, off-balance amount, default status, arrears |
| 4 | Counterparty-instrument | The link between debtors and instruments — role of each counterparty |
| 5 | Joint liabilities | Where several debtors share one instrument — each debtor’s liability share |
| 6 | Accounting | The accounting classification — impairment stage and amount, carrying amount, write-offs, forbearance status |
| 7 | Protection received | Collateral and guarantees — type, value, valuation approach and date |
| 8 | Instrument-protection received | The link between instruments and protections — allocated value |
| 9 | Counterparty risk | Probability of default of the counterparty (IRB institutions) |
| 10 | Counterparty default | Default status of the counterparty and the date it arose |
3. The identifier discipline
- LEI first — where the counterparty holds one. The same LEI register you maintain for DORA and MiFIR serves here.
- National identifier fallback — each country has a ranked list of acceptable identifiers (trade-register numbers, tax IDs). Use the highest-ranked one available.
- Internal identifiers must be stable over time — contract ID and instrument ID cannot change between months without a documented migration, or the ECB sees a disappearance plus a new instrument.
4. The attributes that hurt
Five clusters cause most first-year data-quality findings:
- Counterparty sector and size classification — the ECB sector taxonomy and the SME size test both need data (turnover, balance-sheet total, headcount) that credit files often lack for smaller debtors.
- Purpose and amortisation type — legacy loan systems rarely store the AnaCredit purpose codes; back-filling is manual.
- Forbearance status — needs a link between the credit-risk workflow and the reporting layer; see our dedicated AnaCredit forbearance piece.
- Protection valuation — collateral values, valuation dates and approaches must reconcile with the credit-risk system, and the instrument-protection allocation has to sum correctly.
- Probability of default — IRB institutions must report the PD consistently with the prudential models; standardised institutions leave it empty rather than approximating.
5. Monthly vs quarterly attributes
Most datasets are monthly. Accounting-linked attributes (impairment, forbearance, carrying amount) may follow the reporting agent’s accounting cycle and are reported quarterly where the institution reports FINREP quarterly. The national central bank’s implementing instructions set the exact split — Bundesbank, Banque de France, DNB and Banco de España each publish their own schedule on top of the ECB manual.
6. The validation layers
- Schema validation — field formats, code lists, mandatory fields.
- Referential integrity — every instrument links to a counterparty record; every protection link points to an existing protection.
- Consistency checks — cross-dataset rules (e.g. outstanding amount vs off-balance amount vs commitment).
- Plausibility checks at the ECB — month-on-month movements flagged for explanation; national central banks pass the questions down.
7. FAQ
How many attributes does AnaCredit require?
Roughly ninety-five across the ten datasets, though the count an individual institution actually populates depends on its business (IRB vs standardised, collateralised vs unsecured, joint liabilities or not).
Is there one file or several?
The logical model is ten datasets; the physical submission format is defined per national central bank. Some collect one combined file, others separate counterparty-reference from instrument data.
Do I report in my accounting currency?
Amounts are reported in the currency of denomination of the instrument, with the national central bank handling euro conversion for aggregation. Follow your NCB’s instructions for the exact convention.
What happens when a debtor merges or changes identifier?
Counterparty reference data is updated with the new identifier and the NCB’s prescribed linking procedure. Silent identifier swaps look like a disappeared debtor plus a new one — a classic plausibility flag.
How does the model relate to FINREP?
AnaCredit is instrument-level; FINREP is aggregate. Impairment stages, forbearance flags and carrying amounts should reconcile between the two at the totals level — supervisors cross-check.
8. What to do, today
- Build the counterparty master first — identifiers, sector, size class — before touching instrument extraction.
- Keep contract and instrument IDs stable across systems migrations; document any re-keying.
- Reconcile forbearance and impairment attributes with the FINREP process, not in parallel to it.
- Automate the four validation layers internally before the NCB sees the file.
Related: AnaCredit for payment firms · AnaCredit phases and timeline · AnaCredit forbearance reporting
