Data Quality Challenges for Financial Institutions and large corporations
BI-Community.org Seminar (V1_6)
30/4/2009, Leuven A ÂŤ BU Corporate Reporting Âť presentation Thibaut De Vylder, Nicolas Sayde, Quentin Deschepper
1
DQ in newspapers
2
DQ in newspapers
3
Cost of DQ in perspective In billion US $
1200 1000 800 600 400
200 0
2009
2010
2011
Estyimated cost of DQ problems for U.S. Businesses
600
600
600
Obama's Plan Jan 2009 Federal Spending
1000
Madoff's Fraud
50
Belgium PIB (2007 base in US$)
452
(*) Source:Data Warehousing Institude, Data Quality and and the Bottom Line: Achieving Business Success through a Commitment to High Quality Data, http://www.dw-institute.com/
4
Objective & Experience
Objective:
Present the DEPFAC “data governance” and “data quality” approaches.
Banking environment relevant experience: Regulatory compliance (Basel 2) & corporate reporting
5 billions of data sourced monthly representing hundred of billions in assets & liabilities Chains supported by old and new systems Non homogeneous IT infrastructure (Mainframe / Server…) Overlapping responsabilities
5
Data Governance definition “Data Governance is a system of decision rights and accountabilities for informationrelated processes, executed according to agreed-upon models which describe who can take what actions with what information, and when, under what circumstances, using what methods.� DGI - Gwen Thomas
Reliable information for right decisions
6
Challenge in assembly lines
2nd hand paper Paste transformation
Wood
Paper
End Product transformation
Paper paste
Boxes
Water
Controls on raw material
Defects
Controls on intermediate products Controls on processes
Controls on final products
7
Management challenge in Financial Institutions Process for creating information
Management Report
data
Management decision
Executives base their management decision on information received
ď ľ
ď ľ
How are data proceeded, checked and cross checked ? Are decisions taken on the basis of reliable management reports ?
8
A Data Governance challenge
Real World
Data are transferred, stored, extracted, prepared, calculated and reconciled several times before being reported? A long and risky journey !
Operational systems
A
Central Chains
t1 tranfer
t2 storing
B
C
t3 extraction
D
t4 preparation
E
t5 calculation
F
t6 reporting
G
Information G in report depends on succession of embedded tranformations = t6(t5(t4(t3(t2(t1(data in operational system A))))))) ďƒ 20 to 30% of data may be lost or deteriorated during the process ! 9
A Data Governance challenge A
t1 tranfert t2 storing
Reality is even more complex Duplication of stores Many chains in parallel High risk reconciliations between chains Human factor Re runs Errors and corrections
B
t3 extraction t4 preparation t5 calculation t6 reporting
C
D
E
F
G
t3’ extraction t4’’ preparationt5’ calculation t6’ reporting
D’ T3’’ extraction t1 tranfert
H
I
F’
G’
T4’’ preparationT5’’ calculationT6’’ reporting
D’’ t2 storing
E’ E’’
F’’
G’’
t3 extraction t4 preparation t5 calculation t6 reporting
J
F
L
M
t3’ extraction t4’’ preparationt5’ calculation t6’ reporting
J’
F’
L’
M’
Real chains look more like this
10
Some Data quality dimensions
Real world
Accuracy
Completeness
Integrity & Bus. Rules
Operational systems
A
Central Chains
B
C
D
E
F
G // Chains
D’
E’
F’ This Month Month - 1 Month - 2 Quarter - 1
Consistency
Consistency
Consistency
Intra-chain
Inter-chains
Cross-Months 11
A Data Quality Factory besides the chain
Local
DQ source data
Central Thermometers & KPI’s
Process Quality
Stress, Sensitivity, Simulation…
Prod Cube
DQ Factory
Back Office
Prevention
Front Office
Analysis
Stress Cube Control
12
DQ Framework for DQ continuous improvement
13
Planning & Resources
14
Data Governance applications
Basel 2 Chains
Already operational (Europe) Being implemented (US, Middle east)
Solvency 2 Chains
Corporate reporting chain in Financial Institutions
Banks Insurance companies Regulators
Any « high data volume » reporting chains: telecom industry, postal services, invoices, travel reservation systems, hospitals…
16
Conclusion Large corporate reporting chains must be supported by a Data Quality factory
One € invested in Data Quality improvement has a greater ROI than any other investment (such as adding additional pieces of
software, redevelopements…)
When budgets are scarce, investment in Data Quality is the best investment strategy
17
Deployments Factory Services
Back and Front Office implementation
DQ Factory & DQ Framework
Reconciliate the past and the present (data and processes) Continuous DQ improvement process
Stress Factory
Automated production of reporting & data quality information Automated analysis and communication
Understand the future (through simulations, stress, sensitivity analysis, capital allocation…)
Bypass
« Do things differently » Re-write the whole chain (process and data) in an integrated and homogeneous environment with fixed price implementation, fast delivery and reduced operating costs
18
Why Deployments Factory?
Unique culture of Data Quality management
Expertise and experience
in Financial Industry with Risk, Finance, IT and various Business Lines In complex non homogenous system environments
Able to deliver short term
International & mobile consultants
methodological & pragmatic approach
« Off the shelf » tools & processes: no additonal IT investment required
Limited budgets for great returns…. … the Best ROI you can get ! 19
APPENDIXES
Appendix 1 : Sample of DQ issues Appendix 2 : DQ issue and challenge Appendix 3 : DQ Quotes
20
AP1 : Sample of DQ issues for Basel 2 Syntaxic Inaccuracy Completeness
Semantic Inaccuracy Counter party ID
Name
Counter party Type
Exposure Clasee
EAD
PD
LGD
Start Maturity
End Maturity
123
SME Trilili
SME
Mortgage
100
1%
NULL
2008
2011
124
Company Coca Coli INC
CORPORAT E
Corporate Fin
120
NULL
30%
2010
2012
125
Company HP INC
SME
Corporate Fin
1000000
NULL
0.45
2007
2024
126
Trululu SME
SME
Mortgage
10000
110%
2500
2007
2024
127
Mr John
INDIVIDUAL
Personal Loan
1000
2%
45%
2006
Intra-relation Integrity
Inter-relation Integrity 21
AP2 : DQ issue & challenge defined
22
AP3 : DQ quotes
Source: http://www.dqguide.com
23