IFR/IFD · Class 2
IFR/IFD prudential reporting for Class 2 investment firms
Calculates the nine K-factors and the own-funds requirement, then produces the filing. Every figure records the window, the coefficient and the article behind it.
Built for euro-area investment firms, starting with France. No client in production yet.
| Permanent minimum capital | 750 000.00 |
|---|---|
| Fixed overheads | 250 000.00 |
| K-factor total | 1 284 500.00 |
| art. 11 binding constraint: K-factors | |
| Daily trading flow, average | 250 000 000.00 |
|---|---|
| Coefficient | 0.1 % |
| Window applied | M-4 → M-9 |
| K-DTF | 250 000.00 |
| art. 33(1), TARGET2 business days only | |
The problem
Regulatory reporting platforms are built for large institutions and priced for them. Smaller firms manage with inadequate tools, and with solutions that do not fit.
In November 2025 the FCA published a data quality review of prudential reporting across some 3,800 UK investment firms. Four years into the regime, firms were still leaving the K-factor requirement blank on their returns, and around 20% reported an own funds threshold requirement that did not comply with the guidance. IFR/IFD is the continental equivalent of that regime.
These are rarely typing errors. They come from a calculation applied to the wrong set of days, a window shifted by a month, or a figure that passes every format check because no check looks at whether it is correct.
How it works
Five stages, from your files to the filing. Each one can be inspected.
- 01Your data
- 02Measurement and classification
- 03Calculation engine
- 04Templates
- 05xBRL-CSV package
Deterministic
The same inputs under the same rule and taxonomy versions produce the same figure. A filed return can be re-run years later and will reproduce to the cent.
Traceable
Every amount records the components it was built from, the window applied, the coefficient used and the article it comes from. That detail is available on screen for each figure in the return.
Versioned on two axes
Regulatory rules and EBA taxonomies change at different times. They are versioned separately, so a taxonomy update does not touch the calculation engine. An archived filing keeps the versions it was computed under.
What it does
The nine K-factors, with the measurement windows taken from the regulation.
- K-AUM
- K-CMH
- K-ASA
- K-COH
- K-DTF
- K-TCD
- K-CON
- K-NPR
- K-CMG
The regulation sets out two exclusion conventions, one for monthly data and one for daily data. Both are implemented separately, as written.
The own-funds requirement
The highest of permanent minimum capital, fixed overheads and the K-factor total. The result names the constraint that binds.
Reads your file as it is
Your column names, your delimiter, your date format. The mapping is proposed, you confirm it, and it is saved for the next upload.
Errors you can act on
Each error gives a stable code, the line it occurred on, and the correction needed.
Maker-checker
A preparer cannot approve their own run. Approval freezes it: hashed, timestamped and read-only.
An evidence file
Inputs, versions, results and the full calculation trace, exported as one file for a supervisor or an auditor.
The filing itself
The EBA templates are populated and packaged as xBRL-CSV, then validated against the taxonomy before you submit.
On AI
Calculations come from the engine. The model is used only to explain them.
A language model is used for two tasks on this product, and neither of them affects a reported figure.
No figure submitted to a supervisor comes from a language model. The calculation is deterministic and auditable. In prudential reporting, a plausible wrong number does more damage than a visible failure.
The two tasks are matching the column names in your file to the expected data model, and turning a detected anomaly into a sentence the person fixing it can read. The application runs the same way with that layer switched off.
Current limits
This is an early product. What it does not cover today:
| Out of scope | Own-account activity | Firms dealing on own account, underwriting or operating a trading venue are out of the initial scope. There is no trading book support and the related templates are not built. The product targets agency firms at the moment. |
|---|---|---|
| Out of scope | Class 1 and Class 3 | Class 2 only. Class 3 (small and non-interconnected) and Class 1 are not covered. |
| Euro area | Jurisdiction | The engine generates packages for any euro-area regulator, each on its own national filing calendar. Non-euro EU states are out of scope: they report in national currency, on a different business-day calendar. |
| Provided input | K-NPR | Market risk follows the CRR framework. The figure is taken from your own market-risk system and is not recalculated. |
| Manual | Filing | The package is produced for you to upload to the portal. There is no direct submission yet, but we will be working on it. |
| No pilot yet | In production | No client is running it in production. If you are considering being the first, talk to us before you commit to anything. |
Where the data will live
- Tenancy
- One isolated instance per client
- Region
- European Union
- Filed runs
- Immutable, kept for the retention period
- Compliance frame
- DORA, as an ICT third-party provider
One instance per client, with no shared database. As an ICT third-party provider, DORA requires incident notification, audit rights and a documented file for your outsourcing register.
None of this is live yet: there is no client and no production instance. This is the intended deployment model, set out so you can assess it in advance. The public demo is hosted differently, as described below.
Demo
A working instance with synthetic data. Sign in with the shared account below.
Reset every night. Do not upload real data.
The first request takes about 30 seconds to wake the instance.
It is a demo, which is why it holds no real data. Production runs single-tenant in the EU.
Calculation runs inside the request, so a large run takes a few seconds.
User demo@ifrep.eu
Password demo
No approval, no xBRL-CSV package, no taxonomy validation.
Free validator
Check a report package against the real taxonomy. No account.
EBA RF 4.2 IF modules, xBRL-CSV packages (.zip). Not xBRL-XML, not other frameworks.
Three packages per address per day, 20 MB each.
Processed in France. Deleted as soon as it is read, before validation starts. Results deleted when you fetch them.
The taxonomy's 1264 documents are checked, not just your file.
A preview, not a submission receipt. A receiving portal may apply further checks.
Where it comes from
ifrep is built from inside the industry that supplies regulatory reporting software to large institutions: the EBA frameworks, the data point models and the XBRL taxonomies those platforms run on.
Class 2 firms file under the same regime and read the same articles, without any of that tooling at their scale. Closing that gap is what this product is for.
Get in touch
If you file these returns, we would like to hear how you do it today: what breaks, what takes the longest, and what you have stopped trying to automate.
That is useful to us at this stage, whether or not you would ever use the product.
No newsletter and no automated follow-up. We reply personally.