Tutorial on Matrix Schema Michele Romanelli (BdI)

Matrix Schema Tutorial
Presented at the:
IX European Banking Supervisors
XBRL Workshop & Tutorial
In: Paris
On: 29th September 2008
By: Michele Romanelli
Member of CEBS XBRL Network
Agenda
Introduction – What is a matrix schema
Change management approaches
A deeper view on matrix schemas
Conclusions
The problem: how to align business and
technical view
what business people ask
for
How to guarantee consistency
among requirements, taxonomy
and documentation ?
what technical people say they’ve done
what technical people deliver
What is a Matrix Schema?
• In COREP & FINREP contexts: A method to visualize a complex
multidimensional taxonomy
• Based on a “matrix” where:
 Row headers are the “concepts” to be reported
 Column headers are the involved “dimension”
 Cells determine: (i) if the dimension applies to the concept and
(ii) what are the allowed/disallowed values for that dimension
An Example of a matrix schema
Row headers:
Primary items
Column headers:
Dimensions used in
the template
Each row describes
the breakdown
structure of a
primary item
Each cell contains a
code that stands
for the list of
members allowed
for the dimension
in the primary
item.
A sheet for each
“extended link”
An Example of domain “sheet”
Main advantages
• Very “compact” format
• Useful as a tool to verify the “matching” between business
requirements (normally expressed through templates) and
technical implementation (XBRL taxonomy)
• understanding of COREP/FINREP taxonomies by supervised
entities (hopefully) easier.
• Easy to build and to understand; for COREP & FINREP, it is
automatically derived from XBRL taxonomies through a reverseengineering software.
Taxonomy creation: the current path
From the business template
(presentation centric view)…
… to the taxonomy
(technical view) …
… to the matrix schema
(data centric view):
Reconciliation
Taxonomy creation: an alternative path
User view:
both business template and
corresponding draft matrix schema
Taxonomy building:
no technical experts’ interpretation of
business requirements
Matrix schema derived from the
taxonomy
Reconciliation based on comparison
between input and output matrix
schemas
Change Management approaches
In the near future both FINREP and COREP need to be
amended.
•Change the template only and build the taxonomy from
scratch?
•What about a different approach?
Template vs. Matrix schema
• Templates are mainly concerned with the presentation
layer
• Matrix schemas are mainly concerned with the data
definition layer
The harmonisation should be sought at the data level
Matrix model: a layered, conceptual tool
Transformations
Data structures
Taxonomy 1
Reusable
Concepts
Taxonomy 4
Taxonomy 2
Taxonomy 3
Concepts in matrix model
•
•
•
•
Domains (or Sets)
– Example: the set of European Cities, or the set of Exposure
Classes
Members (or domain-members)
– Example: a single European city (Paris, Rome, …) or a single
Exposure Class (Institutions, Corporates, …)
Dimensions
– A specific meaning of a Domain (example: the city of birth, the
city of residence, the city of work)
Measures (or primary-items)
– aspects we want to observe
Data Structures:
Relationships among concepts
Primary
Item
Is observed through
Domain
Dimension
Is a subset of
hierarchies
Member
Sub-Domain
Building the Matrix schema
1. Think about what you want to observe: Primary Items
2. Think about break-down structures you are interested
in: Dimensions
3. Ask yourself if these dimensions are new ones or
existing ones; assign a name only to new ones;
4. Ask yourself which is the “general domain” of these
new dimensions: is it an existing one? Do you need a
new one?
Building the Matrix schema /2
5. Define members for newly created domains
6. Link appropriate dimensions to the primary item
7. Ask yourself if, in that relationship, the dimension is
allowed to assume all the values of its “general
domain”; if not, identify or define a sub-domain
8. Link the dimension to the general domain or one of its
sub-domains as appropriate
Building the Matrix schema: complex issues
•
Disallowing specific combinations
•
Reporting a primary item for two (or more) different
break-down structures
•
Hierarchies of primary items (just for presentation?)
•
Hierarchies of members (just for presentation? Or
necessary to specify roll-up/drill-down operations?)
•
Modelling tables with a non-evident distinction between
measures and dimensions
Example of disallowed combinations and multiple
break-down structures
Example of tables with non-intuitive distinction
between measures and dimensions
Conclusions
•
Matrix schema as a simple, conceptual tool
•
Suitable to be automated
•
Good chance to automatically generate the taxonomy
from the matrix schema
•
No reconciliation needs
•
Shorter time-to-market
Questions?
Thank you
www.c-ebs.org
www.corep.info
www.finrep.info
Michele Romanelli
E-mail: [email protected]
+39 06 4792 6218
The XBRL Network of the