COREP - FINREP - XBRL SGCB COREP - FINREP - XBRL The French Banking Commission approach VII European Banking Supervisors XBRL Workshop Munich, 09-10 may.

COREP - FINREP - XBRLSGCB

COREP - FINREP - XBRLThe French Banking Commission

approach

VII European Banking Supervisors XBRL Workshop Munich, 09-10 may 2007

Sylviane DELARUEJerome POUPARDCOFINREP Project

Secrétariat général de la Commission bancaire


2


COREP- FINREP- XBRLThe French Banking Commission approach

1 • Project objectives : a european framework

2 • Consequences on our IT system

3 • Challenge : managing complex taxonomies

4 • The French Banking Commission COFINREP

project

5 • COFINREP : technical choices - XBRL learning

curve

3


1. Project objectives : a european framework

• The objectives of Committee of European Banking Supervisors :

– Harmonising banking supervision practices in Europe, and developing a common supervisory culture

– Reducing unnecessary costs of reporting to the banking supervisors through European convergence

• In order to fulfill this goal, need for :

– The definition of a common framework

– A common vocabulary based on the European legislation

– A degree of flexibility in order to adapt to national specifics.

4


• Two european initiatives :

– A common reporting for a common solvency ratio (COREP – COmmon REPorting)

CRD Directives 14 June 2006: final adoption of Directives 2006/48/CE et

2006/49/CE

COREP 13 January 2006: CEBS publishes COREP templates =>

http://www.c-ebs.org/standards.htm December 2005/ juin 2006 : Discussion with the banking industry

on the French reporting

Adoption of the French COREP project by the Banking Commission 26 March 2007


5


– A common « supervisory financial » reporting for banks consolidated accounts in IFRS (FINREP, FINancial REPorting)

European regulation introducing IFRS is directly applicable in Europe

European reporting published by the CEBS on 16 December 2005 => http://www.c-ebs.org/standards.htm

Adoption of the French FINREP project by the Banking Commission on 28 June 2006

1st remittance date for COREP and FINREP 30 June 2007 in France


6


• XBRL taxonomies built by the CEBS for COREP and FINREP are essential in order to create a common framework. They include :

– A common vocabulary for European supervisors, elaborated on the basis of official European regulations and

– A common « grammar » through its « Linkbases ».

• XBRL allows for a european flexible approach, including national specific elements

• But XBRL standard must also keep pace with the evolution of business requirements, and the complexity of COREP and FINREP reportings.


7



• The French Banking Commission decided to make it mandatory for the banks to transmit their COREP and FINREP reports, using the XBRL standard for reasons of efficiency and cost.

• The working relation between banks and the BC will change only marginally :

The General Secretary of the Banking Commission has provided the banks with the adapted French version of the European taxonomies

The financial institutions will transmit their reports (instance documents according to XBRL vocabulary) XML/XBRL, via the usual data transmission means

After verification of data quality, anomalies will be restituted to each institution

8



• The Banking Commission decided to go further in the adoption of XBRL/XmL, for the following reasons : First, our present data system, BAFI, is a proprietary

one, only used for the exchange of data between the French financial institutions and the Banking Commission/Bank of France

BAFI is now a mature system (1993) which was due to evolve in the coming years (very complex legacy in mainframe technology : >1000 programs)

Communication with many applications inside and outside and a much parameterised system (single window betwen financial institutions and BOF).

XmL has become a standard in data transmission which imposes itself in the banking industry

9



• In a first stage, need to run two parallel information systems : COFINREP and BAFI

• Possible since : Relatively few common data between the new

reportings and the present data base

COREP and FINREP are new and would have, in any case, needed important and complex developments in the old system

This situation will not hinder the supervisors’ work

• However, we will need to run data exchanges between both systems

10


2 • Consequences on our IT systems

FinancialInstitution

Information System

Prudential and accounting Reporting

Bâle II

IFRS

COREP Instances

FINREP Instances

SIGNECB

Insideapplications

CB/BDFAccreditation

Input Output

Civil statuts data

Configuration

Accreditation

DataControl

andmanagement

Datamining

XBRLTaxonomy management

BAFI

COFINREP

11



• The second stage (after COFINREP) : we intend to move on with the rest of the reporting system and to have a single data exchange system with the financial institutions.

• In order to achieve this, we will have to :

– Create taxonomies regrouping all necessary data for supervisors

– Add the necessary functionnalities to the new system

– Do the same thing with the other data managed by BAFI (statistics data, monetary data….).

12



• Banking Commission decided to create and manage in-house the French versions of COREP and FINREP taxonomies

• COREP: about 800 entities concerned• FINREP: about 80 entities concerned

• Final users : 280 supervisors and inspectors

• Team of 3 taxonomy managers (and 2 experts)

13



French version of COREP

project considered too complex and too detailed30 templates70.000 data

19 templates14000 data

with 1100 data with a possible postponement 8300 data asked in standard approach and 6000 in IRBA approach

14



French version of FINREP

The CP06, published in april 2005, considered too complex and too détailedThe final version of CEBS (12/2005) took care of these criticismsThe french version (06/2006) reduce even more the reporting

29 templates (France) 1.500 data

39 templates(CEBS)2.400 data

with 400 data (9 templates) yearly 1100 data (20 templates) half yearly

15



• Creation of the French versions on the basis of the European taxonomies

• All the labels of concepts coming from the European taxonomies translated into French

• Creation of a limited number of French specific concepts (with an English and French label)

• Internal validation of the taxonomies with the help of our tools (formalism and standards) and our supplier

16


European taxonomies were available at the end of september for FINREP and on the 6th of october for COREP:

French versions : FINREP => version for discussion (0.1) published on the

Banque de France website on 26 october 2006

- version 0.2 15 january 2007

- approved version 1.0 : 5 february 2007

http://inbdf/fr/supervi/supervi_banc/reporting/finrep.htm COREP => version for discussion (0.1) published on 1st

december 2006

- version 0.2 15 january 2007

- approved version 1.0 : 5 february 2007

http://inbdf/fr/supervi/supervi_banc/reporting/corep.htm


17



Taxonomy editing tool delivery, getting started

and training

T

A

X

O

N

O

M

Y

First draftEuropean taxonomies 01/2006

Launch

Taxonomy extensions

Final version european Taxonomies & Beginning of French taxonomies09/2006

French taxonomy definitions Validated

taxonomies

Taxonomiesavailable for theprofessional public10/2006

02/2007

Project management and supervision

Taxonomy finalization

Professional’s Feed back

18


4. The Banking Commission COFINREP projectSCOPE

– French taxonomies management (VERSIONING)

– Data reception management

– Electronic signature of the instance documents

– Authorisations and reference data management

– Controls of received instances

– Reminders in case of anomalies or late data transmission

– Synthesis and reporting according to the needs of the supervisors

– Interface with the others systems of BDF

– Automatisation of the processes

PORTAL

V1 P

ORTAL

V2

19


4. The Banking Commission COFINREP project

Global conception

Launch

06/2006

Portal V.1

Project management and supervision

Validation

07/2006

Go livePortal V1

Go livePortal V2

P

O

R

T

A

L

Portal V.2

01/200806/2007

20


4. The Banking Commission COFINREP project

• Choice of a generic software :– very short deadlines– use XBRL as far as possible

• Legal process for open tender :– Open tender : 7 december 2005– Emission of specifications : 3 february 2006– Choice of solution : may 2006

• Chosen software meets 50% of users’ requirements in the basic version

• 50% of specific developments taken in charge by the supplier

• Supplier team (5 to 10 people), internal IT team (4 to 6)

21


5 • COFINREP : technical choices - XBRL learning curve

Main challenges :

• Dealing with very complex taxonomies, using dimensions extensively (new standard 1.0 created COREP)

• Lack of maturity of existing softwares, in particular taxonomy editors, not always compatible

• Training offer is still very limited

• Lack of formulae standard, which makes it compulsory to develop in-house business rules

• The question of “versioning” management is urgent and should be a priority for the months ahead

22


The added XML header: why?

Electronic signature

<fichier>

< enveloppe>

<XBRL> <XBRL> <XBRL>

< enveloppe> < enveloppe>

+ To be able to sign multiple instances

+ So that habilitation testing are not XBRL dependent

+ For better « on the fly » performance on instance reception

- Added consistency tests Note: the XML schema of the header is available at:

http://www.banque-france.fr/fr/supervi/supervi_banc/reporting/reporting.htm


23



The workflow of an instance

Definition: an instance in COFINREP, is an XBRL instance limited to a single entity, a single taxonomy, single date and single money (iso421U7).

Definition: The private taxonomy is an extension of our public taxonomy with added controls non compatible with all the existing tools, thus they cannot be shared. These « added » controls are described in the Excel files on the BDF website.

24




During the day:

1. Reception of files

2. If signed, control of the signature and rights to sign

3. Control of the header format (control that all the data is declared production OR simulation in a same file). Implicit XML well-formedness check.

4. Check of the identity of the sender

File received Is it signed? NOCheck XML andheader format

Check identity ofsender

Check signatureand rights to sign

the data

YES

Can the data beextracted form the

signed file?

YES

NO

25




During the night, for every file that passed control 3 during the day:

1. The files are broke down into XBRL instances with a xml header

2. we take the latest file for each key (entity, date, taxonomy)

3. Full validation of instance (public taxonomy) and added controls

4. If no errors, validation with private taxonomy.

5. The instance will then be used for: extraction to other application, transverse reports

6. report to the financial agents on the status of the instances sent

Valide fileBreakdown intoinstances. Takethe latest ones

Control andstorage ofinstances

Blockingerrors?

Control withprivate taxonomy

NO

YES

Transversereports. Extraction

to otherapplication

Report to thefinancial agents

26



# Contrôle Process stop Production Simulation

1In case of a COREP instance, check that the value of the option in the header is properly set to « consolidated » or « social » yes X X

2 XBRL 2.1 yes X X

3 Aggregations and dimensions yes X X

4 Calculation links no X X

5 Uniqueness of key (entity, date, taxonomy) in instance yes X X

6

Coherence of entity between header and instance yes X X

Existence of entity in database Manual verification X X

7

Coherence of date between header and instance

yes X Existence of date in database

8

Coherence of taxonomy between header value and instane’s schemaref yes X X

Validity of the taxonomy and the reported date yes X X

9

Coherence of currency between header value and instance declaration (iso4217 only) yes X

Validity of currency for the entity (database). yes X

10

Is type COREP or FINREP expected for entity

Manual verification X COREP only: is entity authorized for the option declared in the header (consolidated, social)

COREP only: is entity authorized for the approach value in the header, and the values present in the instance. (standard, foundation, advanced).

Manual verification X

COREP only: is the date of the reporting in the instance coherent with the entity’s status - Mars, June, September and December for large groups- June and December only for small groups

Focus on: Control and storage of an instance

27


- Taxonomy published

- Intances workflow

- Taxonomy creation

- Instance management InstanceDocument

BACKEND

Taxonomy

Financials institutions (submitters)

Databasesservers

Corporate Directory Server: BAFI …

InstanceDocument

InstanceDocument

InstanceDocument

InstanceDocument

Receipt Management

Reception ManagementTaxonomies

INTRANETEXTRANET

Taxonomy control

TaxonomyMistakes/Errors

Taxonomy management Control and

storageAmounts ctrl.

BAFI/COFINREPComparision

Reporting synthesis

• Conception• Reporting

calculation

ReportingExtraction

COFINREPReportingErrors

Reporting synthesisReference data management

Habilitations

Taxonomy

InstanceDocument


28



A. The test platform: a compatibility check

The platform opened up in the beginning of March.

• Within one month we received 4 instances

• Instances were mostly FINREP

Result :

• 3 instances: 100% public taxonomy valid. The last instance failed taxonomy validation, but was successful on the second emission

• 3 out of 4 had no CL error, but only one succeeded the private taxonomy tests.

29



B. The performance test

2 0

185

79

5

123

0

20

40

60

80

100

120

140

160

180

200

Step 1 Step 2 Publique taxonom yvalidation

Private taxonom yvalidation

CL logs pars ing Inter-ins tance validation(COREP)

Corep

Finrep

Total

Processing times (minutes)

• Done with over 800 instances.

• Instances with every fact combination filed. Not a “real life” scenario.

• Difficulty encountered: evaluate the target performance.

30



Step 10%

Step 20%

Publique taxonomy validation47%

Private taxonomy validation18%

CL logs parsing1%

Inter-instance validation (COREP)34%

31



Virtual memory usage (On Java batches)

• In a possible future: switching from java 32 bits to java 64 bits (over 4G jvm).

0

100000

200000

300000

400000

500000

600000

700000

800000

900000

1000000

1100000

1200000

1300000

1400000

1500000

1600000

1700000

1800000

1900000

2000000Max allocation memory max used memory Average alloction memory Average used memory

32



CR IR B

21%

M K R SA E QU

11%

M K R SA T DI

11%CA

10%

FINR E P

8%

CR SA

7%

Other

32%

CRIRB MKRSAEQUMKRSATDI CAFINREP CRSAOther

Time consuming factors:

• Number of facts in the instance

• Typed dimensions

• Explicit dimension with many domain-members

33



C. Questions to answer before building your IT

• Precision and decimals

(9453 decimal= -3 )+ (9454 decimal= -3) + (9453 decimal= -3) = 27 023 decimal=-3

We imposed decimal=0 for monetary Items and decimal=4 for pures

Need for a tolerated variation in CL, more precise than the use of decimals (needed: +- 3000)

• Units and scale

Although most existing system work with Kilo-Euros in the BDF, money scale is not part of the XBRL.

We work with Euros, but will have to convert when exporting data to other existing application.

34



• Missing/duplicate facts

Missing facts can make certain Calculation Links to no be check (especially with total)

Duplicated facts will interfere with calculation links, and possibly any control or restitution framework

This is why, it was decided that missing or duplicate fact were to be detected by COFINREP. The financial agents will be warned of these problems, and will be asked to correct this.

•Instance’s size impact

Long item names, namespace ids, unit ids, dimension name….

increase size of instances

Increase processing time of an instance

For example: replacing all ids/names in an instance by numbers => more than 60% size gain

35



• Added control framwork

The need of:

- inter-instance controls

- controls with other data source (data bases, flat file…)

- conditional or complex controls

- Periodic controls (A test only done a particular month or frequency)

Might require to build some configurable control mechanisms.

•Automated restitution: functional/technical barrier on data errors

Not everyone understands fact names, contexts, dimensions and hypercubes.

We need to translate, for example, a Calculation error in way that an accounting personnel will easily understand.

36



• XBRL instances and XML mapping tools

The problem encounter with some XML mapping tools and XBRL instances is often due to the declaration of item:

XBRL instance:

Xsl transformation of XBRL instances is in study

Easier mapping format:

37



D. Xades signature and XBRL

• First test with Xades signature and XBRL instances were conclusive

• Found an incompatibility between XBRL and the c14n, exclusive version, not standard, algorithm. It does not keep silent namespaces.

• The final technical note will soon be available on the Banque de France website

• In the upcoming weeks, the digital signature will be added to our test platform.

38



E. In a nearby future:

We are trying to extend the use of XBRL in other applications:

communication purpose

data processing purpose

Also we follow the interest that major software companies have in XBRL: Oracle, Microsoft, Altova, Fermat…

Ultimately , we communicate with the profession in hope that they will take the XBRL step, not just the export mapping/converter choice!


QUESTIONS ?

Secrétariat général de la Commission bancaire

SGCB

COREP - FINREP - XBRL SGCB COREP - FINREP - XBRL The French Banking Commission approach VII European Banking Supervisors XBRL Workshop Munich, 09-10 may.

Documents

corep finrep

xbrl sgcb corep finrep

xbrl sgcb slide

french corep project

french finrep project

ce corep

ifrs finrep

xbrl taxonomies