数据管理得心应手的10大秘诀
作者： DMR
责任编辑： 阚智
来源： 《电脑商情报》
时间： 2006-09-05 05:37
n an ideal world, the role of the data steward is a critical business function within the corporation and their most powerful tool is a strong master data management strategy. The data steward is the champion for aligning the systems with how the business is managed and yet, within most companies, this function is handled at an application level by a technical data modeler or DBA who just gets the system working. Instead of associating master data with the wealth of information that might be available within the corporation to explain and describe this data, the technical support team makes sure the value is in the table and added to the required hierarchies.
We all sometimes forget that these codes and hierarchies actually define the way in which the business is defined and the way analytics can be performed within the corporation. This is no minor support function, but is a critical strategic function which defines how the business is managed. It is the data steward manager that is ultimately responsible creating the infrastructure required to perform analytic reporting.
In the real world, master data typically has very limited visibility within the business community, and it is not considered a high priority. Not only is this information hidden, but many users maintain their own copy of the data and unstructured data or knowledge relating to the members of a dimension are safely tucked away in a spreadsheet or document on their laptop. Changes to the dimensional hierarchies are processed by the application support team and are requested only when some definition can not be adjusted in the presentation sent to management. There is no shared information and no collaboration on the management of this most critical business asset.
In the ideal world, every company would embrace a robust master data management strategy for all aspects of its business. All the manual effort associated with fixing the data for presentations would be eliminated, information could be shared, information would be auditable and productivity would soar. Analysts could spend time performing analysis rather then gathering and adjusting reports. Reporting would become more reliable and consistent, resulting from the improved data quality; the SOX audit would be much less painful; and the business community would gain more control over the maintenance of this most critical resource.
What is the Value Proposition?
Imagine an executive meeting where sales and marketing are presenting their results to the new president of the division. Sales reports that the market share for a key product line is 25 percent while marketing reports market share at only 15 percent. When questioned by the newly appointed president as to why the numbers are so different, no one in the room could answer the question. How does anyone make a business decision related to market share when no one can explain how two different departments have such dramatically different numbers? Which market share is the correct value? How much effort and resources should the president apply to increasing market share for this product line?
The answer to this puzzle has to do with master data management. The definition of the products contained within the market for the sales and the marketing departments were dramatically different. From a sales perspective, they were targeting very specific products to maximize the sales process and highlight their competitive advantage. However, the marketing team wanted to consider a larger set of products in order to expand the market and increase sales beyond the traditional market. This analysis took several months to understand because although they knew that the reports had come from two different data sources the logic for the market definition was embedded in code that only the systems support group had access to. This caused a significant delay in realigning the marketing and sales strategy for this product. How much this cost the company is difficult to determine, but the sales director was reorganized into another position shortly after this meeting.
What if this company had a master data management strategy that allowed an analyst to determine the definition of a market and compare the difference within two different definitions of the same market by two different organizations? How much time and money would the company have saved? How much additional revenue would the company have generated if a decision to realign the sales and marketing strategy three months earlier? Would the sales director still be in charge of the sales group?
After this incident the company undertook a master data management initiative. They built core master systems for product/market, three types of customers, geography and sales reps/employees. Through this initiative, which was part of a larger data warehousing initiative, they integrated sales and marketing data for the first time in their history. They were also able to derive additional unexpected cost savings from this initiative. They identified that much of the external data that was purchased from third-party vendors was simply a summarized version of the data that they already had at a much more detailed level. As a result, they were able to save costs and replace the feeds to their compensation system, targeting reports, contracting system and the financial systems data file, all because they took the time to define and integrate their dimensional data into a master data management strategy.
If the cost savings and improved decision-making capability aren't enough to convince you that master data management is critical to the business, there is also a legal and fiduciary requirement for ensuring that the definitions of critical metrics used to manage the business are definable and auditable. Sarbanes-Oxley requires a company to be able to explain the data they use to manage the business and report to the public as to the health of the organization. What would have happened to this company if it had been found that they were reporting external results that were not only misleading, but not explainable or auditable all because they didn't actively manage and control their master data?
What is Master Data Management?
When we think of systems in the corporate environment usually we describe the transactional data that these systems are built around. Plenty of attention is given to the transactional data, but the reference data (master data) is ignored until something goes wrong with a transactional report or until two department heads can't answer the questions for the president of the division. The volume of records for master data is typically much smaller than transactional data and unlike transactional data, changes rather slowly. As a result there seems to be less of an urgency to provide a structured solution to maintain this data. When you consider the lack of business priority, the complexity of external data sources and the need to share this information throughout the corporation, the incentive to undertake the development of a core master system becomes less appealing. However, the return on investment (ROI) for implementing a master data management strategy is clearly high. It would be of value even if it only allows all departments within the corporation to consistently interpret the information they get without spending three months of analysis to find the answer to a critical strategic question, but a master data management strategy can yield so many other benefits.
In an ideal world, a master data management strategy includes data (structured and unstructured), process, workflow, collaboration, definition, maintenance facilities, documentation and distribution mechanisms. Master data is a shared resource across all of the systems within the enterprise visible to the business managers. It includes a robust interface to support the browsing of dimensional reference data as well as comparisons of various code structures. The data structures are completely exposed to the business users to help them understand the structure of the technical and business resources within in the company. The user is able to build and maintain their own summarized view of the dimensional hierarchy which can then be utilized in other processes as well as support the management, maintenance, collaboration and documentation features. Implementing a robust master data strategy can yield tremendous benefits to the business and provides not only a high ROI but actually lowers the total cost of ownership (TCO) of other systems that leverage this data.
The Roadmap to Successful Master Data
Building an efficient and effective master data management strategy is a serious commitment that can be a huge task with tremendous benefits, and we have learned by experience that the following principles, if adhered to, will help ensure a successful implementation:
Data Model - The data model is the universal translator and is the foundation of the master data management strategy. The data model of any core master system has certain characteristics in common, such as flexibility, extensibility, controllability and aggregation. The data model has a flexible design that allows for extension of the model to include new sources of data and additional codes. The model has the ability to cross-reference all sources of dimensional data and support all of the detail demographic and profiling data of the code sets. It supports the definition of relationships between the codes, unstructured data and aggregation at both a standard and user defined level.
Historical Archiving - The core master system tracks changes over time in order to provide an audit trail as well as support time-variant analysis. The logic for implementing this functionality conforms to the slowly changing dimension (SCD) logic as defined by Ralph Kimball. The method of implementing the SCD logic is consistent with the requirements of the organization for how to maintain history and combines different types of SCD logic for different parts of the data model.
Aggregation and Hierarchies - The core master hierarchy supports unlimited number of groupings or levels and uneven hierarchies. In some companies, there is a need to have multiple redundant groupings which must be handled appropriately to prevent double counting in the reporting interfaces. Most importantly the model supports a data driven control structure to allow the business to define the rules for creating and maintaining the definition of the dimensional data but still provide the flexibility to change those rules on a global or hierarchical basis.
Workflow - A simple workflow engine is used for implementing functions that require an approval process such as the cleansing process and the maintenance of hierarchical structures.
Data Quality - A key component to any system especially a data warehouse, yet the cleansing process is usually not visible to the business users who are the main beneficiaries of the process. The quality of the data improves when the users participate in the process. The core master systems provide a simplified interface and exchanges files with the cleansing process so that automation of the cleansing process as it involves the business users can be achieved. The business users review the results from the cleansing process and take action to impact the quality of the cleansing process. A historical archive of the results of the cleansing process is maintained to provide a full audit trail. Typically, it is a best practice to implement a three stage cleansing process including: assured automated matching, matching requiring review and informational matching. The key is to involve the business users in the data quality process.
Unstructured Data - There are many pieces of unstructured data or documents that relate to the core master system that can enable a holistic understanding and an audit trail. There are requirements documents, budgets and business plans that can be very valuable to the user community. The master data management strategy stores and relates unstructured data to the code structures within the core master systems.
Access to Information - Providing access to business users is a critical success factor for an effective master data management strategy. The interface is ideally a portal based solution that provides users the ability to manage the code structure and hierarchy, but also to manage the related information about the dimensional structures of the business. A portal becomes a subject- centric application built around dimensional members or groupings. The application provides the ability to browse information about the dimensional reference data, but would also allow the creation of collaborative intelligence. For example, an account management portal provides a single centralized place to keep information that characterizes each account. Of course, you can see groupings of accounts and all of the codes that can represent that account, but you can store and share notes on the account, contacts within the account, contract information and other structured and unstructured information. The application could even provide integrated analytics about the accounts by imbedding the analytics tool right into the portal. It is the portal that empowers the users to take an active role in the management of the master data strategy.
Collaboration - As part of the user interface, it is critical to enable and empower the users to create notes, have threaded discussions, document best practices or create relationships about the structure of a dimension or the codes. This collaboration will create invaluable knowledge for the corporation that can describe the technical structures within the company in terms of the business priorities and give you control over the code sets that are used to define your technical data resources.
Distribution and Data Sharing - In order to ensure that the systems that are dependent upon the master data management process satisfy the needs of the user community, there needs to be a method to distribute changes made within the core master systems. Typically it is a best practice to separate the maintenance of the master data from the usage of the data structures in reporting systems. This gives you the ability to modify code structures without impacting reporting results from either a performance or consistency perspective.
SOA - Developing the core master systems utilizing a service-based architecture allows a system to make a Web services request for information or updates and enables a truly distributed and shared architecture that can help implement a real-time analytics environment. It serves as an alternative method for distribution and data sharing.
We all sometimes forget that these codes and hierarchies actually define the way in which the business is defined and the way analytics can be performed within the corporation. This is no minor support function, but is a critical strategic function which defines how the business is managed. It is the data steward manager that is ultimately responsible creating the infrastructure required to perform analytic reporting.
In the real world, master data typically has very limited visibility within the business community, and it is not considered a high priority. Not only is this information hidden, but many users maintain their own copy of the data and unstructured data or knowledge relating to the members of a dimension are safely tucked away in a spreadsheet or document on their laptop. Changes to the dimensional hierarchies are processed by the application support team and are requested only when some definition can not be adjusted in the presentation sent to management. There is no shared information and no collaboration on the management of this most critical business asset.
In the ideal world, every company would embrace a robust master data management strategy for all aspects of its business. All the manual effort associated with fixing the data for presentations would be eliminated, information could be shared, information would be auditable and productivity would soar. Analysts could spend time performing analysis rather then gathering and adjusting reports. Reporting would become more reliable and consistent, resulting from the improved data quality; the SOX audit would be much less painful; and the business community would gain more control over the maintenance of this most critical resource.
What is the Value Proposition?
Imagine an executive meeting where sales and marketing are presenting their results to the new president of the division. Sales reports that the market share for a key product line is 25 percent while marketing reports market share at only 15 percent. When questioned by the newly appointed president as to why the numbers are so different, no one in the room could answer the question. How does anyone make a business decision related to market share when no one can explain how two different departments have such dramatically different numbers? Which market share is the correct value? How much effort and resources should the president apply to increasing market share for this product line?
The answer to this puzzle has to do with master data management. The definition of the products contained within the market for the sales and the marketing departments were dramatically different. From a sales perspective, they were targeting very specific products to maximize the sales process and highlight their competitive advantage. However, the marketing team wanted to consider a larger set of products in order to expand the market and increase sales beyond the traditional market. This analysis took several months to understand because although they knew that the reports had come from two different data sources the logic for the market definition was embedded in code that only the systems support group had access to. This caused a significant delay in realigning the marketing and sales strategy for this product. How much this cost the company is difficult to determine, but the sales director was reorganized into another position shortly after this meeting.
What if this company had a master data management strategy that allowed an analyst to determine the definition of a market and compare the difference within two different definitions of the same market by two different organizations? How much time and money would the company have saved? How much additional revenue would the company have generated if a decision to realign the sales and marketing strategy three months earlier? Would the sales director still be in charge of the sales group?
After this incident the company undertook a master data management initiative. They built core master systems for product/market, three types of customers, geography and sales reps/employees. Through this initiative, which was part of a larger data warehousing initiative, they integrated sales and marketing data for the first time in their history. They were also able to derive additional unexpected cost savings from this initiative. They identified that much of the external data that was purchased from third-party vendors was simply a summarized version of the data that they already had at a much more detailed level. As a result, they were able to save costs and replace the feeds to their compensation system, targeting reports, contracting system and the financial systems data file, all because they took the time to define and integrate their dimensional data into a master data management strategy.
If the cost savings and improved decision-making capability aren't enough to convince you that master data management is critical to the business, there is also a legal and fiduciary requirement for ensuring that the definitions of critical metrics used to manage the business are definable and auditable. Sarbanes-Oxley requires a company to be able to explain the data they use to manage the business and report to the public as to the health of the organization. What would have happened to this company if it had been found that they were reporting external results that were not only misleading, but not explainable or auditable all because they didn't actively manage and control their master data?
What is Master Data Management?
When we think of systems in the corporate environment usually we describe the transactional data that these systems are built around. Plenty of attention is given to the transactional data, but the reference data (master data) is ignored until something goes wrong with a transactional report or until two department heads can't answer the questions for the president of the division. The volume of records for master data is typically much smaller than transactional data and unlike transactional data, changes rather slowly. As a result there seems to be less of an urgency to provide a structured solution to maintain this data. When you consider the lack of business priority, the complexity of external data sources and the need to share this information throughout the corporation, the incentive to undertake the development of a core master system becomes less appealing. However, the return on investment (ROI) for implementing a master data management strategy is clearly high. It would be of value even if it only allows all departments within the corporation to consistently interpret the information they get without spending three months of analysis to find the answer to a critical strategic question, but a master data management strategy can yield so many other benefits.
In an ideal world, a master data management strategy includes data (structured and unstructured), process, workflow, collaboration, definition, maintenance facilities, documentation and distribution mechanisms. Master data is a shared resource across all of the systems within the enterprise visible to the business managers. It includes a robust interface to support the browsing of dimensional reference data as well as comparisons of various code structures. The data structures are completely exposed to the business users to help them understand the structure of the technical and business resources within in the company. The user is able to build and maintain their own summarized view of the dimensional hierarchy which can then be utilized in other processes as well as support the management, maintenance, collaboration and documentation features. Implementing a robust master data strategy can yield tremendous benefits to the business and provides not only a high ROI but actually lowers the total cost of ownership (TCO) of other systems that leverage this data.
The Roadmap to Successful Master Data
Building an efficient and effective master data management strategy is a serious commitment that can be a huge task with tremendous benefits, and we have learned by experience that the following principles, if adhered to, will help ensure a successful implementation:
Data Model - The data model is the universal translator and is the foundation of the master data management strategy. The data model of any core master system has certain characteristics in common, such as flexibility, extensibility, controllability and aggregation. The data model has a flexible design that allows for extension of the model to include new sources of data and additional codes. The model has the ability to cross-reference all sources of dimensional data and support all of the detail demographic and profiling data of the code sets. It supports the definition of relationships between the codes, unstructured data and aggregation at both a standard and user defined level.
Historical Archiving - The core master system tracks changes over time in order to provide an audit trail as well as support time-variant analysis. The logic for implementing this functionality conforms to the slowly changing dimension (SCD) logic as defined by Ralph Kimball. The method of implementing the SCD logic is consistent with the requirements of the organization for how to maintain history and combines different types of SCD logic for different parts of the data model.
Aggregation and Hierarchies - The core master hierarchy supports unlimited number of groupings or levels and uneven hierarchies. In some companies, there is a need to have multiple redundant groupings which must be handled appropriately to prevent double counting in the reporting interfaces. Most importantly the model supports a data driven control structure to allow the business to define the rules for creating and maintaining the definition of the dimensional data but still provide the flexibility to change those rules on a global or hierarchical basis.
Workflow - A simple workflow engine is used for implementing functions that require an approval process such as the cleansing process and the maintenance of hierarchical structures.
Data Quality - A key component to any system especially a data warehouse, yet the cleansing process is usually not visible to the business users who are the main beneficiaries of the process. The quality of the data improves when the users participate in the process. The core master systems provide a simplified interface and exchanges files with the cleansing process so that automation of the cleansing process as it involves the business users can be achieved. The business users review the results from the cleansing process and take action to impact the quality of the cleansing process. A historical archive of the results of the cleansing process is maintained to provide a full audit trail. Typically, it is a best practice to implement a three stage cleansing process including: assured automated matching, matching requiring review and informational matching. The key is to involve the business users in the data quality process.
Unstructured Data - There are many pieces of unstructured data or documents that relate to the core master system that can enable a holistic understanding and an audit trail. There are requirements documents, budgets and business plans that can be very valuable to the user community. The master data management strategy stores and relates unstructured data to the code structures within the core master systems.
Access to Information - Providing access to business users is a critical success factor for an effective master data management strategy. The interface is ideally a portal based solution that provides users the ability to manage the code structure and hierarchy, but also to manage the related information about the dimensional structures of the business. A portal becomes a subject- centric application built around dimensional members or groupings. The application provides the ability to browse information about the dimensional reference data, but would also allow the creation of collaborative intelligence. For example, an account management portal provides a single centralized place to keep information that characterizes each account. Of course, you can see groupings of accounts and all of the codes that can represent that account, but you can store and share notes on the account, contacts within the account, contract information and other structured and unstructured information. The application could even provide integrated analytics about the accounts by imbedding the analytics tool right into the portal. It is the portal that empowers the users to take an active role in the management of the master data strategy.
Collaboration - As part of the user interface, it is critical to enable and empower the users to create notes, have threaded discussions, document best practices or create relationships about the structure of a dimension or the codes. This collaboration will create invaluable knowledge for the corporation that can describe the technical structures within the company in terms of the business priorities and give you control over the code sets that are used to define your technical data resources.
Distribution and Data Sharing - In order to ensure that the systems that are dependent upon the master data management process satisfy the needs of the user community, there needs to be a method to distribute changes made within the core master systems. Typically it is a best practice to separate the maintenance of the master data from the usage of the data structures in reporting systems. This gives you the ability to modify code structures without impacting reporting results from either a performance or consistency perspective.
SOA - Developing the core master systems utilizing a service-based architecture allows a system to make a Web services request for information or updates and enables a truly distributed and shared architecture that can help implement a real-time analytics environment. It serves as an alternative method for distribution and data sharing.
