Data Modelingfor the BusinessDonna Burbank, CA ERwin
Agenda• Data Modeling for the Business   – Starting with a High-Level Business View   – Tips for Implementing your High-Le...
Who am I?• Donna Burbank has more than more than 15 years of experience in the areas of  data management, metadata managem...
Who Are You? Survey• How would you describe your role?          A. Data Architect, Data Modeler, or Analyst          B. Bu...
The Challenge• You’ve been tasked to assist in the creation of a Data Warehouse• Trying to obtain a single view of ‘custom...
What is a High-Level Data Model?• A high-level data model (HDM) uses simple graphical images to  describe core concepts an...
“A Picture is Worth a Thousand Words”Examples of High-Level Data Models PAGE 7
“A Picture is Worth a Thousand Words”Examples of High-Level Data Models            Product                    Location    ...
“A Picture is Worth a Thousand Words”Examples of High-Level Data Models PAGE 9
“A Picture is Worth a Thousand Words”Examples of High-Level Data Models PAGE 10
“A Picture is Worth a Thousand Words”Examples of High-Level Data Models              UML PAGE 11
“A Picture is Worth a Thousand Words”Examples of High-Level Data Models              UML                                  ...
“A Picture is Worth a Thousand Words”Examples of High-Level Data Models PAGE 13
“A Picture is Worth a Thousand Words”Examples of High-Level Data Models PAGE 14
Is Notation Important?• Many Notations can be used to express a high-level data model• The choice of notation depends on p...
Levels of Data Models PAGE 16
Levels of Data Models• Models can be built  – Top-Down  – Bottom-Up  – Using a Hybrid Approach  PAGE 17
What is in a Name?• The High-Level Data Model goes by many names PAGE 18
Many names for a High-Level Data Model PAGE 19
How is this Different from a Logical Model?                  VHDM                                        HDM              ...
Building a High-Level Data Model• Let’s go back to our challenge, to achieve a ‘single version of the  truth’ for Customer...
Building a High-Level Data Model• We start with a very simple HDM, with just one object on it,  called “Customer”.• We use...
Too simple?• Our team thought so, so went ahead and focused on the  technical integration, including:   – Reverse engineer...
Focusing on the Business• This implementation went “perfectly”, with no errors in the  scripts, no data type inconsistenci...
Focusing on the Business• Until we showed the report to the business sponsor:  – We can’t have 2000 customers in this regi...
Back to the Drawing Board• After doing an extensive review of the six source systems, and  talking with the system owners ...
Oops!• We were mixing current customers, with prospects (non-  customers).  – We just sent a discount coupon to 1600 of th...
Achieving Consensus• We created a report of the various definitions of customer> And verified with the various stakeholder...
Resolving Differences• Our new high-level data model looked like this: PAGE 29
Identify Model Stakeholders• Make sure ALL relevant parties are involved in the design process  – Get buy-in!  PAGE 30
Identify Model Stakeholders• Make sure ALL relevant parties are involved in the design process  – Get buy-in!  PAGE 31
A HDM Facilitates Communication• A High-Level Data Model Facilitates Communication between  Business and IT  – Focus on yo...
Communication is the Main Goal                    of a High-Level Data Model• Wouldn’t it be helpful if we did this in dai...
Some Creative Ways to Facilitate Conversations withStakeholders• Food!   – “Lunch and Learn”   – Bring candy to meetings• ...
Example from ERworld Case Study:Scott Northrup’s Implementation at Wells Fargo  PAGE 35            © 2010 Wells Fargo Bank...
Identify Model Purpose• Key to success of any project is finding the right pain-point and  solving it.• Make sure your mod...
Managing the Technical InfrastructureWhy do you need a modeling tool, and not a drawing tool?• Recall that we had multiple...
The CA ERwin® Modeling Family PAGE 38   *The mark of Saphir is used with the consent of Silwood Technology, Limited
Creating a Data Inventory withCA ERwin Data Modeler• CA ERwin Data Modeler can reverse and forward engineer from all leadi...
Design Layers Create both Business                 and Technical Designs  Business              Data               DBA  Sp...
A Data Model can be your Filter• A Data Model can add:   – Focus – by Subject Area, by Platform, etc.   – Visualization – ...
Filter Information by Subject Area• Subject Areas help you filter by Subject / Content PAGE 42
Create Different Displays for Different Audiences• Stored Displays help you filter by Audience - Business PAGE 43
Create Different Displays for DifferentAudiences• Stored Displays help you filter by Audience - Technical PAGE 44
How ERwin helps Share Information with Various Audiences• Metadata Bridges allows you to import export information from a ...
Inferring Legacy Structures withCA ERwin Data Profiler• For our IDMS system, there is no relational structure or Primary/F...
Understanding ERP Systems withCA ERwin Saphir Option• We’ve now gotten an inventory of our databases: both legacy and  tra...
Understanding ERP Systems withCA ERwin Saphir Option• Using the CA ERwin Saphir Option, we can easily group tables by  sub...
Managing the Data Inventory withCA ERwin Model Manager• CA ERwin Model Manager provides a single repository to store all o...
Reporting – Sharing Information with Stakeholders• Now that we’ve created an inventory of all of our data sources  and man...
Using Crystal Reports for End Users• Many users want to see definitions, but not read a data model.PAGE 51
Case Study – Major International Oil CompanyPAGE 52
Corporate Culture• A diverse, federated organization – culture encourages local decision making  within a corporate framew...
Repurpose “Models”• In addition to using traditional “data models”, the team translated their high-  level data models to ...
Repurpose Models – MS Excel PAGE 55
Community of Interest• The data architecture team created a “Community of Interest” to:   – Share best practices inside co...
Case study lessons• Understand roles and motivations and work within the  organization    –      Federated governance mode...
Case study lessons• Understand roles and motivations and work within the  organization    – Federated governance model    ...
Summary• A high-level data model can help achieve a “single view of  customer”• Aim the HDM at the business user by being ...
Data Modeling for the Business• Available at:   – Amazon.com   – Technics Publications  The authors of Data Modeling for t...
Questions? PAGE 61
Upcoming SlideShare
Loading in …5
×

Data modeling for the business 09282010

1,703 views

Published on

Published in: Technology, Business
0 Comments
1 Like
Statistics
Notes
  • Be the first to comment

No Downloads
Views
Total views
1,703
On SlideShare
0
From Embeds
0
Number of Embeds
18
Actions
Shares
0
Downloads
119
Comments
0
Likes
1
Embeds 0
No embeds

No notes for slide

Data modeling for the business 09282010

  1. 1. Data Modelingfor the BusinessDonna Burbank, CA ERwin
  2. 2. Agenda• Data Modeling for the Business – Starting with a High-Level Business View – Tips for Implementing your High-Level Model – Using CA ERwin solutions – Case Study PAGE 2
  3. 3. Who am I?• Donna Burbank has more than more than 15 years of experience in the areas of data management, metadata management, and enterprise architecture. – Currently is the senior director of product marketing for CA’s data modeling solutions. – Has served in key brand strategy and product management roles at Computer Associates and Embarcadero Technologies and as a senior consultant for PLATINUM technology’s information management consulting division in both the U.S. and Europe. – Has worked with dozens of Fortune 500 companies worldwide in the U.S., Europe, Asia, and Africa and speaks regularly at industry conferences. – Has recently co-authored a new book entitled Data Modeling for the Business, with Steve Hoberman & Chris Bradley PAGE 3
  4. 4. Who Are You? Survey• How would you describe your role? A. Data Architect, Data Modeler, or Analyst B. Businessperson or Business Analyst C. DBA or Technical IT D. A combination of the above E. Other PAGE 4
  5. 5. The Challenge• You’ve been tasked to assist in the creation of a Data Warehouse• Trying to obtain a single view of ‘customer’• Technical and political challenges exist – Numerous systems have been built already—different platforms and databases – Parties cannot agree on a single definition of what a ‘customer’ is• Solution: Need to build a High-Level Data Model PAGE 5
  6. 6. What is a High-Level Data Model?• A high-level data model (HDM) uses simple graphical images to describe core concepts and principles of an organization and what they mean• The main audience of a HDM is businesspeople• An HDM is used to facilitate communication• It needs to be high-level enough to be intuitive, but still capture the rules and definitions needed to create database systems. PAGE 6
  7. 7. “A Picture is Worth a Thousand Words”Examples of High-Level Data Models PAGE 7
  8. 8. “A Picture is Worth a Thousand Words”Examples of High-Level Data Models Product Location Customer Region Order Raw Material Ingredient PAGE 8
  9. 9. “A Picture is Worth a Thousand Words”Examples of High-Level Data Models PAGE 9
  10. 10. “A Picture is Worth a Thousand Words”Examples of High-Level Data Models PAGE 10
  11. 11. “A Picture is Worth a Thousand Words”Examples of High-Level Data Models UML PAGE 11
  12. 12. “A Picture is Worth a Thousand Words”Examples of High-Level Data Models UML ORM PAGE 12
  13. 13. “A Picture is Worth a Thousand Words”Examples of High-Level Data Models PAGE 13
  14. 14. “A Picture is Worth a Thousand Words”Examples of High-Level Data Models PAGE 14
  15. 15. Is Notation Important?• Many Notations can be used to express a high-level data model• The choice of notation depends on purpose and audience• For data-related initiatives, such as MDM and DW: – ER modeling using IE (Information Engineering) is our choice of notation (i.e. “crow’s feet”) – It is important that your high-level model uses a tool that can generate DDL, or can import/export with a tool that can – A repository-based solution helps with reuse and standards for enterprise-wide initiatives PAGE 15
  16. 16. Levels of Data Models PAGE 16
  17. 17. Levels of Data Models• Models can be built – Top-Down – Bottom-Up – Using a Hybrid Approach PAGE 17
  18. 18. What is in a Name?• The High-Level Data Model goes by many names PAGE 18
  19. 19. Many names for a High-Level Data Model PAGE 19
  20. 20. How is this Different from a Logical Model? VHDM HDM LDMDefines the scope, audience, context for Defines key business concepts and their Represents core business rules and datainformation definitions relationships at a detailed levelMain purpose is for communication and Main purpose is for communication and Provides enough detail for subsequentagreement of scope and context agreement of definitions and business first cut physical design logicRelationships optional. If shown, Many-to-Many relationships OK Many-to-Many relationships resolvedrepresent hierarchy.Cardinality not shown Cardinality shown Cardinality shownNo attributes shown Attributes are optional. If shown, can be Attributes required and all attributes are composite attributes to convey business atomic. Primary and foreign keys meaning. defined.Not normalized (Relational models) Not normalized (Relational models) Fully normalized (Relational models)Subject names should represent high- Concept names should use business Entity names may be more abstractlevel data subjects or functional areas of terminologythe businessSubjects link to 1-M HDMs Many concepts are supertypes, although Supertypes all broken out to include sub- subtypes may be shown for clarity types‘One pager’ Should be a ‘one pager’ May be larger than one pageBusiness-driven Cross-functional & more senior people Multiple smaller groups of specialists involved in HDM process with fewer IT. and IT folks involved in LDM process.Informal notation ‘Looser’ notation required – some format Formal notation required construct needed, but ultimate goal is to be understood by a business user< 20 objects < 100 objects > 100 objects PAGE 20
  21. 21. Building a High-Level Data Model• Let’s go back to our challenge, to achieve a ‘single version of the truth’ for Customer information• We have 6 different systems with customer information in them: – 2 on Oracle – 1 on DB2 – 1 using legacy IDMS – 1 SAP system Oracle – 1 using MS SQL Server Oracle DB2 IDMS SQL Server SAP PAGE 21
  22. 22. Building a High-Level Data Model• We start with a very simple HDM, with just one object on it, called “Customer”.• We use an ER Model and show business definitions Too Simple?? PAGE 22
  23. 23. Too simple?• Our team thought so, so went ahead and focused on the technical integration, including: – Reverse engineering a physical model from each system – Creating ETL scripts – Migrating the data into a single hub – Building a reporting system off of the data PAGE 23
  24. 24. Focusing on the Business• This implementation went “perfectly”, with no errors in the scripts, no data type inconsistencies, no delays in schedule, etc.• We built a complex BI reporting system to show our upper management the results.• We even sent out a welcome email to all of our customers, giving them a 50% off coupon, and thanking them for their support. PAGE 24
  25. 25. Focusing on the Business• Until we showed the report to the business sponsor: – We can’t have 2000 customers in this region! I know we only have around 400! – Why is Jones’ Tire on this list? They are still evaluating our product! Sales was negotiating a 10% discount with them, and you just sent them a 50% coupon!?!? – You just spent all of that money in IT to build this report with bad data??? PAGE 25
  26. 26. Back to the Drawing Board• After doing an extensive review of the six source systems, and talking with the system owners we discovered that: – The DB2 system was actually used by Sales to track their prospective “customers” – These “customers” didn’t match our definition—they didn’t own a product of ours!! PAGE 26
  27. 27. Oops!• We were mixing current customers, with prospects (non- customers). – We just sent a discount coupon to 1600 of the wrong people! – We gave upper management a report showing the wrong figure for our total number of customers! – We are now significantly over budget to have to go back and fix this!!• We started over, this time with a High-Level Data Model PAGE 27
  28. 28. Achieving Consensus• We created a report of the various definitions of customer> And verified with the various stakeholders that:  There were 2 (and only 2 definitions) of customer  Sales was OK with calling their “customer” a “prospect” PAGE 28
  29. 29. Resolving Differences• Our new high-level data model looked like this: PAGE 29
  30. 30. Identify Model Stakeholders• Make sure ALL relevant parties are involved in the design process – Get buy-in! PAGE 30
  31. 31. Identify Model Stakeholders• Make sure ALL relevant parties are involved in the design process – Get buy-in! PAGE 31
  32. 32. A HDM Facilitates Communication• A High-Level Data Model Facilitates Communication between Business and IT – Focus on your (business) audience • Intuitive display • Capture the business rules and definitions in your model – Simplicity does not mean lack of importance • A simple model can express important concepts • Ignoring the key business definitions can have negative affects – A model or tool is only part of the solution • Communication is key • Process and Best Practices are critical to achieve consensus and buy-in PAGE 32
  33. 33. Communication is the Main Goal of a High-Level Data Model• Wouldn’t it be helpful if we did this in daily life, too?• i.e. “Let’s go on a family vacation!” Person Concept Definition Father Vacation An opportunity to take the time to achieve new goals Mother Vacation Time to relax and read a book Jane Vacation A chance to get outside and exercise Bobby Vacation Time to be with friends Donna Vacation More time to build data models PAGE 33
  34. 34. Some Creative Ways to Facilitate Conversations withStakeholders• Food! – “Lunch and Learn” – Bring candy to meetings• Force? – “No bathroom breaks until we reach consensus!”• Active Listening – Understand why there is disagreement (e.g. “Ingredient” vs. Raw Material)• Fit into their schedule – Webinars – The “5 minute rule” for business execs – small, bite-sized models or questions. PAGE 34
  35. 35. Example from ERworld Case Study:Scott Northrup’s Implementation at Wells Fargo PAGE 35 © 2010 Wells Fargo Bank, N.A. All rights reserved.
  36. 36. Identify Model Purpose• Key to success of any project is finding the right pain-point and solving it.• Make sure your model focuses on a particular pain point, i.e. migrating an application or understanding an area of the business Existing Proposed Business “Today an Account can “By next quarter, an only be owned by one Account can be owned by Customer.” more than one Customer.” Application “In the legacy Account “When we migrate to Management system, we SAP/R3, Account Holder call the customer an will be represented as Account Holder.” Object.” PAGE 36
  37. 37. Managing the Technical InfrastructureWhy do you need a modeling tool, and not a drawing tool?• Recall that we had multiple data sources on a variety of platforms: – 2 on Oracle – 1 on DB2 – 1 using legacy IDMS – 1 SAP system – 1 using MS SQL Server• How can CA ERwin help manage this? Oracle Oracle DB2 IDMS SQL Server SAP PAGE 37
  38. 38. The CA ERwin® Modeling Family PAGE 38 *The mark of Saphir is used with the consent of Silwood Technology, Limited
  39. 39. Creating a Data Inventory withCA ERwin Data Modeler• CA ERwin Data Modeler can reverse and forward engineer from all leading DBMSs – we use this to inventory our Oracle, DB2, and SQL Server data sources• “Design Once, Reuse Many Times” across heterogeneous platforms• Design layers allow you to have a single high-level model pointing to numerous physical model platforms. Oracle DB2 SQL Server PAGE 39
  40. 40. Design Layers Create both Business and Technical Designs Business Data DBA Sponsor Architect Physical Data Model Logical Data Model (Oracle) (Business Area 1)Conceptual Data Physical Data Model Model (SQL Server) Logical Data Model (Business Area 2) Physical Data Model (DB2) PAGE 40
  41. 41. A Data Model can be your Filter• A Data Model can add: – Focus – by Subject Area, by Platform, etc. – Visualization – Different Views for Different Audiences – Translation – to different DMBS AND to non DBMS formats such as UML, BI tools, Excel, XML, etc, etc. CA ERwin Oracle Oracle DB2 Developers Business Sponsors ETC! ETC! DB2 SQL 3NFServer IDMS SAP Data Architects DBAs PAGE 41
  42. 42. Filter Information by Subject Area• Subject Areas help you filter by Subject / Content PAGE 42
  43. 43. Create Different Displays for Different Audiences• Stored Displays help you filter by Audience - Business PAGE 43
  44. 44. Create Different Displays for DifferentAudiences• Stored Displays help you filter by Audience - Technical PAGE 44
  45. 45. How ERwin helps Share Information with Various Audiences• Metadata Bridges allows you to import export information from a variety of sources: ETL, BI tools, ER modeling tools, UML tools, MDM hubs etc. PAGE 45
  46. 46. Inferring Legacy Structures withCA ERwin Data Profiler• For our IDMS system, there is no relational structure or Primary/Foreign keys defined.• We need to infer the structure from actual data values using CA ERwin Data Profiler, so that we can import them into CA ERwin Data Modeler to reuse as part of our enterprise architecture. PAGE 46
  47. 47. Understanding ERP Systems withCA ERwin Saphir Option• We’ve now gotten an inventory of our databases: both legacy and traditional DBMS. But what do we do about our SAP system? – There are thousands of tables – When we reverse engineer them, we get unintuitive German technical names PAGE 47
  48. 48. Understanding ERP Systems withCA ERwin Saphir Option• Using the CA ERwin Saphir Option, we can easily group tables by subject area, and can translate table and column names into intuitive, English versions.• And can more easily integrate SAP data models into our enterprise data architecture. PAGE 48
  49. 49. Managing the Data Inventory withCA ERwin Model Manager• CA ERwin Model Manager provides a single repository to store all of your data model assets• A collaborative environment for multiple modeling teams.• Metadata storage for: multiple models, multiple dbms platforms, multiple tools, multiple audiencesMultiple Multiple Multiple Tools MultipleModels DBMSs Audiences Oracle Teradata BI Tools DB2 SQL Developers Business Server Spreadsheets ETL Tools Sponsors Single Definition of 3NF “Customer” CA ERwin Model Manager Data Architects DBAs PAGE 49
  50. 50. Reporting – Sharing Information with Stakeholders• Now that we’ve created an inventory of all of our data sources and managed them with central models, we want to share this information with users across the company.• CA ERwin is now bundled with Crystal Reports to generate reports for both business and technical users such as – Logical/business attribute definitions – Physical data structures• In addition, the ERwin ODBC interface allows you to create similar reports using other tools: Cognos, Pentaho, Excel, Access, etc. PAGE 50
  51. 51. Using Crystal Reports for End Users• Many users want to see definitions, but not read a data model.PAGE 51
  52. 52. Case Study – Major International Oil CompanyPAGE 52
  53. 53. Corporate Culture• A diverse, federated organization – culture encourages local decision making within a corporate framework• Before high-level models were introduced: • Data architecture performed in different Segments & Functions • Variety of tools & techniques used • Projects encountered common cross-business data concepts, but largely created their own models & definitions • No overall context for models existed • Negative image regarding the term “models” PAGE 53
  54. 54. Repurpose “Models”• In addition to using traditional “data models”, the team translated their high- level data models to ⁻ Excel Spreadsheets ⁻ Word Documents & Reports ⁻ HTML Pages on the web• It was the same information, but translated into a format the business users could understand. PAGE 54
  55. 55. Repurpose Models – MS Excel PAGE 55
  56. 56. Community of Interest• The data architecture team created a “Community of Interest” to: – Share best practices inside company – Exchange ideas across projects• The goal was to get ALL users involved via – “Lunch and Learn” meetings – Webinars – Training and Education• Both Business and Technical resources were invited PAGE 56
  57. 57. Case study lessons• Understand roles and motivations and work within the organization – Federated governance model – Avoid silo mentality – Communicate – Start small & document success – Make it easy to get hold of – Market, market, market!• Follow up with a robust architecture – Common repository – Models appropriate for the audience – Defined ownership/stewardship – Unique definitions – “Repurpose” data for various audiences: via the web, Excel, DDL, XML, etc. It’s the data that’s important, not the format. PAGE 57
  58. 58. Case study lessons• Understand roles and motivations and work within the organization – Federated governance model – Avoid silo mentality – Communicate – Obtain buy in by starting small & document success – Make it easy to get hold of – Market, market, market!• Follow up with a robust architecture – Common repository – Defined stewardship – Unique definitions – “Repurpose” data for various audiences: via the web, Excel, DDL, XML, etc. It’s the data that’s important, not the format. PAGE 58
  59. 59. Summary• A high-level data model can help achieve a “single view of customer”• Aim the HDM at the business user by being “generic” but keeping enough detail to make it meaningful• Treat your model like a project – Identify pain points and solve them – Identify stakeholders and market – Document purpose and expected results – Follow and organized, repeatable process• Using high-level models can help increase communication with the business and achieve better results PAGE 59
  60. 60. Data Modeling for the Business• Available at: – Amazon.com – Technics Publications The authors of Data Modeling for the Business do a masterful job at simply and clearly describing the art of using data models to communicate with business representatives and meet business needs. The book provides many valuable tools, analogies, and step-by-step methods for effective data modeling and is an important contribution in bridging the much needed connection between data modeling and realizing business requirements. Len Silverston, author of The Data Model Resource Book seriesPAGE 60
  61. 61. Questions? PAGE 61

×