Title: Using MDM as a Practical Approach to Get Started in Data Governance Todd Goldman: VP Products and Marketing
1Using MDM as a Practical Approach to Get Started
in Data Governance Todd Goldman VP Products
and Marketing
2Precursor to Data Governance is Data Management
3Data Governance Manifesto
- Data should be
- Understood
- Secure
- Consistent
- Accessible
- Managed
Data Governance
4Governance Requires a Foundation of Understanding
- If you dont know how data in different systems
is related How can you make sure they are
consistent (MDM)? - How can you measure overall data quality?
- How can you measure the quality of business
rules? - If you dont know where the sensitive data
is How can you protect it? - If your data is not secure, consistent and
accessible What does it mean to manage it?
5Even if you understand your data
landscapestarting data governance is
difficult.It is a new religion.
6Current Data Religion Mystery Cult
- Data is shrouded in mystery
- Its meaning is only accessible through data
priests - Data Priests use omens (metadata) and personal
experience to divine the meaning of data in
return for (financial) sacrifices - Meaning is often obscure, misleading, incomplete
and wrong - Data Priests are often blockers to data
governance programs - More paperwork
- Slows us down
- Dont need it
7A Common Myth We know our data
Im a professional. Of course I know my data!
- Subject matter experts (SMEs) only know their own
systems - But they cant tell you how it changes and is
transformed as it moves from system to system - Relationships between systems are complex
- SMEs sometimes change jobs!
But, once it leaves my hands, it is someone
elses problem!
Wow, that transformation is complex. Are you
sure that is in my data?
Im going to start my own consulting firm
8More of the Myth Our Data is Consistent
All of my data follows the business rules for
this system!
- Business rules are broken all the time as data
crosses business and system boundaries - 83 year old man in system A is a youthful
driver in system B - Bond yield is listed as 5 in system X and 5.3
in system Y - Exceptions result in lost revenue, customer
dissatisfaction, and regulatory fines - Business rules change as organizations change
- Mergers and Acquisitions
- New products or services
- Products/services are retired
- Reorganizations
- New IT systems are added
I cant keep up with all the acquisitions and
reorganizations. They mess up the way systems
work together. It is very inconvenient.
9Data Ecosystem
- Warriors
- Have immediate, acute data problem
- Focus on feasibility and time
- Tradespeople
- Have a short term (project) business problem
- Focus on value
- Data Priests (SMEs)
- Know the data
- Assist with data problems
- Reformers
- Have a long term business problem
- Focus on control and scalability
10Data Governance New Data Religion
- Governed Data
- Data is documented, consistent and secure
- Governance is a must for MDM projects
- To succeed, a new religion needs
- Reformers or prophets people who believe in it
and sponsor it - Priests people to educate, explain and promote
it - New Miracles successful projects and spectacular
results
11Successful Data Governance Program Roll Out
Help Warriors win battles
Use victories to convince Reformers to change
religions
Convert the priests
Have priests drive adoption to trades people
12Key Winning the First Battle!
- Pick your battles
- Find appropriate initial project
- Achieve quick win
- Early success
- Build a Trojan Horse
- Govern the data without calling it that
13Picking The First Battle
- Appropriate Project
- Immediate Business ROI
- Project success directly linked to Data
Governance practice and methodology - Cross-Silo
- Must have for the company
- Examples
- Basel II
- Master Data Management
- Application migration
- Cross-BU Reporting and Analytics
14Avoid False Starts
- Projects to Avoid
- Future ROI next project will benefit
- Boil the ocean
- False Start Examples
- Refactoring
- Metadata repository
- Enterprise (fill in the blank)
15Quick Win
- Iterative Approach
- Agile Development
- Immediate results
- Automation is Critical
- Data Discovery tools
- Repeatability of results
- Validation and consistency tools
- Incident/Exception workflow
- Visual and Intuitive Presentation of Results
- Business oriented
- Graphically presented
16Barriers to the Quick Win
17Data Governance Gap
The Peaks of Data Understanding
Design specifications get lost or outdated,
subject matter experts leave companies, databases
and business rules get changed without updating
documentation, mergers and acquisitions wreak
havoc on databases, all leading to a company not
knowing exactly what they have... The end result
is inconsistent data. Fern Halper, Hurwitz
Associates
Data Governance
Data Nightmare
70 or more of the time and effort involved in
completing most data integration projects is
consumed by defining and implementing the
business rules by which data will be mapped,
transformed, integrated, and cleansed. Ted
Friedman Vice President, Gartner Group
18Current Tools Werent Developed to Discover a
Distributed Data Landscape
- ETL, EAI, Cleansing
- Not discovery solutions. They depend on
discovery - Metadata matching
- Doesnt work in a real environment
- Profiling
- Focused on a single data source
- Todays tools werent created to analyze a
distributed data landscape - Data analysts manually examine data values to
figure out the business rules in the distributed
data landscape - The most sophisticated tools commonly used today
is
Most Widely Used Business Rule Discovery Tools
19Case Study Asset Master
Reminder for Todd Show Dawns slides
Vice President Charlotte, NC based Commercial bank
- Project
- IT Asset Master
- Consolidating 8 asset management systems to a
single asset master
We had 9 subject matter experts spend 9
months and we still didnt know enough to be able
to consolidate our data into a master.
20Data Analysis The Lack of UnderstandingA Case
Study(Note This is NOT what you want to do!)
21Data Analyst Case Study
- The story you are about to hear is true
- Only the names have been changed to protect the
innocent
22This is Denise
- Experienced Data Analyst
- Extremely successful career working for data
software companies - Very Personable
- Very Intelligent
- Impeccable references
- Bills at 2000/day
- Hired by a dental insurance company for a 3 week
data analysis/MDM integration project - Tools used
- Profiling
- TOAD
- SQL
- Highlighter
Data Analyst Denise
23Manual Data Discovery Timeline
- Get metadata specs and begin to check business
rules between one table with six columns against
first of three source systems - Expected result
- 3 Weeks
Day 1
Data Analyst Denise
24Manual Data Discovery Timeline
- Get initial results from unit test with
inconsistent data for 1st column - So far, so good
Day 2
Data Analyst Denise
25Manual Data Discovery Timeline
- Retest and debug
- Still on track
Day 3
Data Analyst Denise
26Manual Data Discovery Timeline
- Go to data architect to question
- Architect pings owner of application (SME).
- NOTE Data analyst not allowed to consult with
SME directly.
Day 4-5
Data Analyst Denise
27Manual Data Discovery Timeline
- Meeting with architect and SME to review.
- Initial answer received .
Day 6
Data Analyst Denise
28Manual Data Discovery Timeline
- Rewrite business rules and test.
- Find second column with inconsistent data.
- Retest and debug.
Day 7
Data Analyst Denise
29Manual Data Discovery Timeline
- Go to data architect to question
- Architect pings owner of application (SME).
- SME asks upstream application owner
Day 8-10
Data Analyst Denise
30Manual Data Discovery Timeline
- Flurry of emails between the 4 players, as
upstream app owner in different time zone. - Decision on how to proceed agreed upon
Data_Architect_at_company.com
Day 11-12
SME_at_company.com
App_Owner_at_company.com
Data Analyst Denise
App_Owner_at_company.com
App_Owner_at_company.com
denise_at_company.com
denise_at_company.com
denise_at_company.com
App_Owner_at_company.com
31Manual Data Discovery Timeline
- Rewrite business rules in SQL and test.
- Find more inconsistent data.
- Retest and debug.
Day 13
Data Analyst Denise
32Manual Data Discovery Timeline
- Go to data architect to question
- Architect pings owner of application (SME).
Day 14
Data Analyst Denise
33Manual Data Discovery Timeline
- Meeting with architect and SME to review.
- Decision made to review specs with a larger group
Day 15
Data Analyst Denise
34Manual Data Discovery Timeline
- Meeting with larger group.
- Original specs validated and corrected
Day 16
Data Analyst Denise
35Manual Data Discovery Timeline
- At weekly status meeting, project manager asks,
why have 17 days passed when this phase was to
be completed in 3 weeks?
Day 17
Data Analyst Denise
36Manual Data Discovery Timeline
Day 18
Data Analyst Denise
37Manual Data Discovery Timeline
- Pass first source system SQL to ETL developers
for coding and QA
Day 19
Data Analyst Denise
38Manual Data Discovery Timeline
- Get specs and begin to verify relationships with
second of three sources systems an outside feed
Day 20
Data Analyst Denise
39Manual Data Discovery Timeline
- Go to data architect to question
- Architect pings owner of application (SME).
- SME asks upstream application owner
- Feed vendor liaison is consulted
Day 21-23
Data Analyst Denise
40Manual Data Discovery Timeline
- Flurry of emails between the 4 players, plus
vendor liaison. - More people involved consumes even more time
- Decision on how to proceed agreed upon
Data_Architect_at_company.com
Feed_vendor_at_company.com
Day 24-26
SME_at_company.com
App_Owner_at_company.com
Data Analyst Denise
App_Owner_at_company.com
App_Owner_at_company.com
denise_at_company.com
denise_at_company.com
denise_at_company.com
App_Owner_at_company.com
41Manual Data Discovery Timeline
- Recode SQL and test.
- Repeat experience of days 7-16, with new
inconsistent data
Day 27
Data Analyst Denise
42Manual Data Discovery Timeline
- Recode SQL and test.
- Repeat experience of days 7-16, with new
inconsistent data
Day 37
Data Analyst Denise
43Manual Data Discovery Timeline
- The project now 18 days overdue, with no clue as
to how long it will take to complete the
remaining work. - Repeat variations of days 21-37 several times
Day 37
Data Analyst Denise
44Manual Data Discovery Timeline
- Pass 2nd source system business rules to ETL
developer and QA. - Project phase is now 70 days overdue, with one
entire source system still to code. - Red flags being raised
- Search for sacrificial lambs.
Day 89
Data Analyst Denise
45Manual Data Discovery Timeline
- Go on preplanned, and much overdue vacation
Day 90
46Manual Data Discovery Timeline
- Get specs and begin to check business rules with
third of three sources systems. - Repeat variation of days 20-89.
Day 97
98
99
100
102
104
106
108
110
120
130
140
150
160
165
166
167
47Manual Data Discovery Timeline
- Pass 3rd source system code to ETL developer and
QA. - Project is 152 days late
- 30 weeks
- 7 months
- Company paid for 30 weeks more consulting time
than expected - 300K overrun
Day
167
48What does this mean for your Data Governance
Project?
49MDM deployment10MM in Services for Every 1MM
in Software
Services
- MDM Hub
- Merge
- Purge
- Match
Software
50According to the Experts
70 or more of the time and effort involved in
completing most data integration projects is
consumed by defining and implementing the
business rules by which data will be mapped,
transformed, integrated, and cleansed.
Ted Friedman Vice President Gartner Group
5170 of Services are for Data Analysis30 of
Services Are for Deploying the Data Hub
Services
Data Analysis Services
MDM Deployment Services
Software
MDM Hub
52According to the Experts
Malcolm Chisholm MDM Industry Expert AskGet, Inc.
MDM wont ever provide a positive return on
investment to businesses if the cost and risk of
the data analysis and mapping component is not
reduced by an order of magnitude you have to
automate the process
53Recap So Far
- You must overcome big hurdles you must overcome
to implement Data Governance
- Technical
- Presumes data understanding
- Requires automated data discovery, validation,
remediation for a distributed data landscape
- Financial
- Cost of deployment must justify the project
- Cultural
- You may be changing religions
54Data Governance Epic
The Peaks of Data Understanding
Data Governance
Data Nightmare
55Data Governance Epic
- But you know the alternatives are unthinkable, so
you and your team of data governance warriors
boldly go where no man has gone before.
56Data Governance Epic
- Scale the cliffs of data relationship discovery
- Pick your way through data inconsistency glaciers
- Battle Data Priests for budget and mindshare
57Data Governance Epic
- And eventually, if you are very, very persistent
and very, very lucky, you may even get there
58Case Study Potential for the False Start
- Manufacturing Firm
- Corporate mandate to improve data quality (the
CEO demanded a new religion) - Created their initial identity master
- Initial Identity Master
- Required 5 analysts to map 4 data sources
- Merge purge match process is governed
- But Quality of data in the master is suspect
- Result No downstream users
- Next project
- 16 more sources to map
- Will require 20 more data analysts
- The Problem
- Hiring 20 data analysts is not financially
feasible - Data mapping and analysis is the critical path
- Millions of dollars have been spent on software
and services already
59Theres got to be an easier way!
Need a Quick Win
60What if you could Automate Cross System Data
Understanding?
61Automating cross system data discovery would
change the economics of governance from this
Services
MDM Deploy- ment Services
Software
MDM Hub
62Automating cross system data discovery would
change the economics of governance from this
Services
MDM Deploy- ment Services
Software
MDM Hub
63Automating cross system data discovery would
change the economics of governance to this
- Provides the foundation of good data management
- Automates understanding of the current data
landscape - Replace services with software (10x differential)
- Creates repeatability
- Makes data governance projects financially
feasible - Accelerate deployment
- Reduce project risk
- Turn negative NPV into positive NPV
- Provides the Trojan Horse
MDM Deploy- ment Services
Software
Analysis
MDM Hub
64Case Study The Trojan Horse
- Truck Manufacturer
- Migrating from one finance application to another
- Data must be mapped and migrated as part of the
process - The Trojan Horse
- Some data in the finance application is master
data - Using automated tools to map the data and will
leverage the map to create a master - Did a pilot project where automation took 3 days
vs 6 months for manual mapping - Planned savings from automation are being
rerouted to purchase an MDM system - Critical Factors
- Governance processes will be required to clean up
the data as part of the migration - They are not calling this governance they are
just doing it - All mapping efforts will be leveragable because
they are repeatable and verifiable - Repeatable and verifiable are good words
- Future Challenges
- They must execute
65Data Discovery Automation TechnologyA Primer
66Automated Cross System Data DiscoveryWhat is it?
- New data analysis methodology and tools
- Arms the warrior with a new weapon
- Allows you to quickly understand your current
data landscape - Establishes data understanding within data
sources and between data sources - Automates discovery of business rules, lineage,
transformations and data inconsistencies across
data sources - Goes well beyond profiling
- Examines
- Data Values
- Data Values
- Data Values
- Establishes a methodology for cross system data
analysis - Each data project becomes a building block, not a
one-off
67Data-Driven Approach Aligns Rows Across Datasets
- Step 1 Discovery Engine analyzes the data values
to automatically discover the key that aligns
rows across disparate datasets - Works for hundreds of tables
- Works for millions of rows
68Data-Driven Approach Aligns Rows Across Datasets
- Step 1 Discovery Engine analyzes the data values
to automatically discover the key that aligns
rows across disparate datasets - Works for hundreds of tables
- Works for millions of rows
Table 1
69Data-Driven Approach Discovers Business Rules
Sensitive Data
- Step 2 With rows now aligned, analyzes the data
values to automatically discover - Forgotten Business Rules
- Data Lineage
- Hidden Sensitive Data
Table 1
70Data-Driven Approach Discovers Business Rules
Sensitive Data
- Step 3 With business rules now discovered,
analyzes the data values to automatically
discover - Unknown Data Inconsistencies
Table 1
71What Complex Business Rules are Discovered from
the Data?
- Scalar
- One to one
- Substring
- Concatenation
- Constants
- Tokens
- Conditional logic
- Case statements
- Equality/Inequality
- Null conditions
- In/Not In
- Conjunctions
- Joins
- Inner
- Left Outer
- Aggregation
- Sum
- Average
- Minimum
- Maximum
- Column Arithmetic
- Add
- Subtract
- Multiply
- Divide
- Reverse Pivot
- Cross-Reference
- Custom Data Rules
72Case Study Worldwide Financial Institution
- Financial Services Firm
- Integration of legacy system with reference
master system. - First of 40 to be integrated
- Manual results for first dataset
- Estimated to take 6 months elapsed
- Data-Driven Mapping results
- 2.5 weeks of elapsed time
- Also centralized data analysis expertise
- Benefits
- Significant time to market savings 5 months
- Significant project risk reduction
- Data inconsistencies found as part of process
Master Data Management (Deployment time)
Months
73What does this all mean?
- Makes it much easier and cheaper to map your
distributed data landscape - This is the foundation upon which the rest is
built - The economics of governance will look very
different - Faster, repeatable victories
- Turns point projects into governance building
blocks - Undoable projects become doable
- Turns data governance projects on their heads
74Recap I
Culture use victories to build the case for
better data governance and quality
Trojan Horse Start governing your data without
calling it governance
Financial Use better data management to deliver
positive ROI
Technical Automate data discovery and management
75One more point about Culture Change
Strategy
Communications
Organizational Structure
Culture
Training/Skills
Rewards
You cant just change culture. You have to turn
other knobs that affect culture
76Recap II
77Data Governance Success
- Happy CXO Management Team
78Questions and Answers
- Thank You for Attending!
- For more information, contact
- Todd Goldman
- Web www.exeros.com
- Email todd_at_exeros.com
- Phone 1.408.213.8910
- Or stop by Exeros Booth in the exhibit hall