This document discusses applying lean principles to data quality management (DQM) and risk/compliance reporting. It describes lean DQM as focusing on reducing waste by minimizing time spent building unwanted solutions through shorter iterations, prototyping, and a build-measure-learn cycle. A lean approach pairs data quality specialists and business users to prototype reports/dashboards and get early, actionable feedback. This validates requirements and priorities before full implementation. The document demonstrates this approach using tools like X88 Pandora and Tableau to prototype risk reporting by financial instrument, region, and industry sector to meet new regulatory requirements.
1. How to Apply Lean DQM principles
for smarter delivery of risk and
compliance reporting
2. Data to Value Ltd. introduction
5 Core practice areas
Lean Information Management
Specialists
Expert users of Technology
accelerators for bridging
technology & business Data gap
Strategy &
Architecture
Governance
&
Management
InsightModelling
Quality
LEAN
INFORMATION
MANAGEMENT
3. What is Lean in the context of IM?
Originated in manufacturing before
adoption in other areas including:
Software Development (Extreme
Programming)
Process optimisation (Six Sigma)
Project Management (Agile, Scrum
etc.)
Business startups
Focus on reducing waste
Key mantra is minimising time
spent on building solutions
customers do not want
Lean
Information
Management
Shorter
iterations
Prototyping
& Minimum
Viable
Products
Build-
Measure-
Learn cycle
Early
adopters
Cross
functional
teams
Actionable
metrics
4. Traditional DQM & report
development approach
Requirements
Design
Implementation
Testing
Structured approach with clearly
delineated stages &
responsibilities
Numerous interfaces &
touchpoints between specialists
e.g. ‘IT’ & ‘the business’
Fully tested ‘product’ delivered at
the end of the project
What happens if:
Requirements change?
Deadlines move?
The customer doesn’t know what they
want?
6. Lean approach - architecture
Real-time
connectivity
Disparate Files
& Databases
analysed
Data Sources
Prototyped reports
& dashboards
- Funds
- Securities
- Clients
- Trades
- Portfolios / Books
- Valuations /
Prices
- …
Cleanse
Parse
Standardise
Validated
feedback from
business users
achieved by
prototypingIssues, Insights and actions fed
back into cycle
Data
Prototyping
Data
Profiling
Data
Modelling
Our approach & toolset is proven at helping clients to extract more value
from existing and prospective datasets
Document
Reverse
Engineer Discover
Transform
7. Lean approach – pairing &
presentation
Data Quality
Management
Data Prototyping
Key performance indicators for
Data Quality SMEs
Bottom up DQ
Prototyped dashboards for
Business stakeholder SMEs
Top down DQ
9. New financial regulations often have
long lead times & undergo many peer
reviews / amendments
No longer enough to simply provide
data - must be able to understand &
articulate end to end process too e.g.
BCBS 239
Financial Markets data landscape
semantically diverse
Firms often leave difficult challenges
to end of process
Use case - risk & compliance reporting
Hypothetical
requirement to report &
demonstrate analysis
of trading risk by:
• Financial
Instrument Type
(Asset Class)
• Geographic Region
• Industry sector
2012 20192016
FATCA / IFRS
AMLD IV
Solvency II
EMIR
UCITS V
10. Passionate, Innovative, Lean
Lean Information Management specialists
93 Western Road
Tring
Hertfordshire
HP23 4BN
T +44 (0) 208 278 7351
www.datatovalue.co.uk
Nigel Higgs
nigel.higgs@datatovalue.co.uk
James Phare
james.phare@datatovalue.co.uk
Notas del editor
1
Cofounder, London-based IM consultancy. Help clients apply Lean IM techniques to maximise value of data & info assets. Specialise primarily in Financial Services but also undertake work in other sectors.
Previous experience – Reuters, Group Head of DA & IM @ Man – one of worlds largest alternative asset managers.
5 core practice areas – Consultants are cross functional (trained in fundamentals of all areas) -- recognition of interconnected nature of discipline.
Lots of disruptive things happening in data space at present e.g. Data Science Hackathons. Appetite to do things in a faster, more iterative & collaborative way. We are interested in lean because we see this as a vehicle for meeting these demands.
Manufacturing Origins
Lots written about – not a new idea, but relatively new to IM field.
Focus on waste, maximising outputs from inputs & minimising non-value add activities. Lots of interest recently with lean startup – applying learnings to tech & other startups.
Examples of waste in IM:
Defects (e.g. data quality issues)
Excessive production / engineering (e.g. developing data architectures, approaches or solutions not demanded by customers)
Over-processing (e.g. implementing overlapping data quality checks, reporting solutions etc.)
Duplication (e.g. maintaining the same data or information in different forms)
Excessive motion (e.g. unnecessary transport / integration of data assets)
Missed insight, opportunity or potential (e.g. missing sales opportunities through incomplete data)
Rework & delays (e.g. repeatedly defining the same datasets and requirements or sourcing the same type of data)
Complexity (e.g. overly complicated data architectures)
Risk (e.g. unnecessary operational, reputational, legal, financial and other risks emanating from poor IM)
Talk through spokes….Designed to minimise this waste – is something we apply to all IM initiatives we undertake, including DQM.
Linear segregated stages & responsibilities. Traditional approach for reporting usually involves data warehouse, data mart or other database delivered at end with reporting tool.
Traditional waterfall approach works well if:
Requirements are clear and relatively static.
Scope & timelines are defined and don’t change.
You have experienced people in place.
People not collocated (suppliers, regions etc)
Customer feedback is not critical
Often doesn’t work so well if quantity of change or uncertainty is increased.
Swing diagram analogy
Lean & Agile approach is much more accommodating of change & less prescriptive.
Iterative, hypothesis driven process.
Utilises cross functional expertise.
Starts with understanding current state – discovering characteristics of data, outliers, hypothesising about issues.
Moves onto capturing actionable metrics (not vanity metrics) and building in business metrics that sponsors care about – cost, risk. Essentially answering the so what question using facts. Number of defects, types of defects, What is true cost of defects, what is risk of defect?
Metrics are presented alongside prototyped deliverables such as in this case reporting deliverables. Helps to achieve ‘validated learnings’ which can be fed back into cycle for next iteration.
Final result is a working prototype that helps to:
Clarify requirements.
Better understand obstacles for full production implementation.
How does this work in practice?
We have developed an architectural pattern using a number of tools to work in an iterative way.
Right architectural components the key – using X88 Pandora & Tableau today. There are other tools that you can apply this approach using however.
Will talk more about the profiling, prototyping & presentation aspects today.
Essentially however using approach able to rapidly load a variety of sources, model & document these, prototype assumptions and deliver working prototype DQ dashboards, reports, insights and other deliverables faster than traditional methods.
7
Why risk & compliance reporting a good use case?
Lots of uncertainty & change – mention timelines.
No longer enough just to provide data. Regulators starting to examine process and hand out fines – e.g. RBS.
Semantics of Data landscape – initiatives like EDM Council FIBO working through common definitions after many others have tried – e.g. SWIFT, FIX, ISO20022.
Firms often look at regs not too differently to how school children view homework. Leave most difficult bits to last. Lean suggests you should try to tackle these first or at least understand them e.g. RUP.
Example uses data that resembles that which an investment manager would use. Example is deliberately straight forward however for non-FS participants.