Master Data Management

Strategy Paper

Prepared by: Elliot Daly

8 July 2011

Contents
Strategy Paper ................................ ................................ ................................ ................................ ................ 1 Contents ................................ ................................ ................................ ................................ ......................... 2 Executive Summary ................................ ................................ ................................ ................................ ........ 3 Vision ................................ ................................ ................................ ................................ .............................. 4 Strategy ................................ ................................ ................................ ................................ .......................... 4 Data governance & quality................................ ................................ ................................ ........................... 4 Data integration services ................................ ................................ ................................ ............................. 5 Metrics ................................ ................................ ................................ ................................ ............................ 6 Governance ................................ ................................ ................................ ................................ .................... 6 Representation................................ ................................ ................................ ................................ ............ 6 Organisation ................................ ................................ ................................ ................................ ............... 7 Team Responsibility................................ ................................ ................................ ................................ .... 7 Deliverables ................................ ................................ ................................ ................................ ................ 8 Technology Architecture................................ ................................ ................................ ................................ 9 Overview................................ ................................ ................................ ................................ ..................... 9 Identifying Systems ................................ ................................ ................................ ................................ ..... 9 Coexistence Architecture................................ ................................ ................................ ........................... 10 Appendix A ................................ ................................ ................................ ................................ ................... 11 Assumptions ................................ ................................ ................................ ................................ ............. 11

Page | 2

"Virtually everything in business today is an undifferentiated commodity, except how a company manages its information. How you manage information determines whether you win or lose." Bill Gates

Executive Summary
In all organisations a huge of amount of information is generated or referenced every day, from unstructured information through to hierarchical, transactional data. This information or data is a corporate asset collectively owned by the entire enterprise. How an organisation manages and utilises this asset will 9 times out of 10 determine the relative success or not of that organisation. Before successfully exploiting this corporate asset we need to put in place a robust foundation;this will only be achieved through the creation, adoption and governance of our organisation s master data. Companies that have managed to do this successfully have been able to drive growth, improve operational efficiency and ease the burden of regulatory compliance. Employee & client data reside in a number of locations throughout our organization from enterprise wide systems through to people s local machines. This impacts the business in a number of ways: 1. Reliability & usability of data effects ability to make appropriate business decisions. 2. Operational efficiencies compounded by duplication of data in different places and high transaction and data processing errors. 3. Corporate data is not leveraged as an asset to provide any competitive advantage. By implementing a Master Data Management (MDM) programme we will aim to mitigate this business impact. This initiative will endeavor to deliver: 1. Leaner operational efficiencies through adoption of better data management practices for both employee and customer master data. 2. Increased revenue by providing a single version of the truth for customer data. By reducing time spent looking for information more effort could be focused on customer engagement. 3. Improved cross and up sell opportunities by providing more accurate customer management information 4. Ability to be more agile in response to our customer s need. 5. Ability to target our customers that are the most profitable. 6. Ability to track all employees across the organisation. 7. Mitigate compliance risks such as employee insurance and protection of intellectual property once an employee leaves. 8. Time saved during the company acquisition process & onboarding. 9. Time saved updating systems after departmental, regional or global re organisation.

The costs associated with poor data can be enormous. According to a recent report titled The Cost of Poor Data Quality by Dun and Bradstreet and the Ivey School of Business (2007), poor data quality costs the U.S. economy six hundred billion dollars annually 5% of the American GDP.

Page | 3

Vision
MDM must underpin the business strategic vision: To have a revenue of at least 200m by 2013 To focus on margin over revenue (implicit efficiency); target for 2011 8.8%, with further improvements thereafter To continue developing (strategic) clients and sectors To further globalise the business and move more towards an enterprise culture, without losing core values.

To deliver on this vision it is recognised that 1. 2. 3. 4. Improved management information is the key to better decision making Driving efficiencies is a permanent business priority that will continue to improve margin Through the One Firm at Work initiative and focusing on the rope strategy we can up and cross sell to both existing and new clients. We need to retain common practices and a single, global operating model

MDM underpins this by providing the process, structure & governance for managing employee, customer and hierarchy enterprise data. Operational efficiencies will be achieved by eradicating duplication of effort, streamlining the flow of information from a centralised data store to consuming systems and reducing the amount of time to find correct information. In addition by providing quick reliable information to sales teams and client facing staff we can look to drive cross sell and up sell opportunities.

Strategy
To realise the vision we recommended delivery of MDM in a phased approach think strategically but deliver tactically. The programme will be split into three distinct phases covering Employee Data Integration (EDI), Customer Data Integration (CDI) & Hierarchy Information Management (HIM). Each phase wil l include two main tenets: data governance & quality and data integration services.

Data governance& quality

A data quality program within a data governance framework is a cross -functional, enterprise-wide initiative requiring that everyone, regardless of their primary role or job function, accept a shared responsibility for preventing data quality lapses and for responding appropriately to mitigate the associated business risks when issues do occur. Data governance not only reveals the business value of the organization's data but also reveals the communication and collaboration necessary to materialize that value as positive business impacts. The enforcement of data governance policies is often confused with traditional management notions of command and control, which is the antithesis of what it takes to successfully implement a data governance program. What data governance really demands is an organizational culture that embodies collaboration. Information Management Magazine 2010

Without reliable controls on our data there would be no benefit implementing any technology solution to aid MDM. MDM starts with governance structures and clean data and <Company Name> will need to embark on a programme of work to deliver this. The root cause of many data quality issues is the lack of a formal enterprise-wide data management strategy and we will need to adopt better data management practices. This will include the creation of a data quality team and associated governance structure with defined and adhered to procedures, standards and policies. Assuming our systems run as they should the main source of bad data is human error. In our environment this ranges from failure to enter data properly through to data inconsistencies and data duplication. Improving data quality will require <Company Name>to manage the full information life cycle from the moment of entry

Page | 4

through to the point when it is no longer in use. This needs to be facilitated by enforceable data entry standards and practices and strong policies on data access and security. We recommend that these standards, practices and policies are signed off by Executive Committee and relevant GPOs and data stewards and data users are accountable for upholding the policy. Please refer to governance section for details on data governance team, the roles and responsibilities.

Data integration services

Data Integration services refers to the technology that underpins the MDM programme. These technology tools vary considerably from best of breed MDM toolsets that focus on a particular data set e.g. customer data to full enterprise MDM tools that provide services that are data agnostic and include extensive additional functions such as cleansing, de-duplication, process workflow and synchronisation. Master data life cycle One of the key elements to delivering a robust technology platform will be to fully understand the life cycle of our master data. Where is it created and by whom? How is the data used across our organisation? At what point does the data get archived or does it need to be retained and for how long? These are some of the key questions that need to be answered before we can understand the requirements that MDM technology needs to support. The responsibility for analysing and mapping this out lies with the data governance committee and data integrity teams (see section on organisation and responsibilities).

Master data architecture Changing our applications to read from a centralised data store is not an approach recommended by experts in the MDM space. Instead it is recommended to separate data management from business logic and transactional systems. To change and make significant changes to our applications will be costly and will elongate the time taken to deliver successful MDM. By creating a separate layer of technology for managing data we can achieve: 1. Lower build costs 2. Lower on-going maintenance costs 3. Faster time to value for other projects 4. More agility and adaptability to business change

Figure 1 A separation of data management from business logic of applications is recommended

Page | 5

Metrics
For this initiative we need to have measurements to measure the success or otherwise of the MDM programme. These should really be focused on business drivers and some of these have been outlined in the vision and executive summary sections. Some of the key metrics that we can tie back to MDM include: Corporate Goals Focus on Margin over revenue Development of an Enterprise Culture Strategic Goals Improved management information for cross sell and up sell opportunities Driving efficiencies leading to improved margin Operational Goals Agility and innovation reduce time to market for IT projects Better business process and stewardship Data Quality

Governance
Representation
Creating a governance structure for data quality management will require input from both business and technical representatives. The business is responsible for communicating the business needs regarding what data is needed, how it is collected by end-users, and how it will ultimately be used by business decision makers. The IT department should provide specialists that will execute on the technical piece of data collection, storage, infrastructure, and integration. In too many cases the IT department focus too heavily on delivering a technical solution to MDM. Delivering effective MDM is about people and processes and adherence to those processes. Business stakeholders will need:
y

y y y

The authority to create and implement enforceable procedures, standards and policies to improve data quality. This authority must be issued from senior management to ensure enterprise-wide compliance. The ability to understand the causes of poor data in the organization with both the authority, motivation, and desire to improve data quality. The capacity, drive and political power to develop and implement and enforce solutions at both the departmental and enterprise level. The capacity to audit and evaluate the success of the solutions and take necessary actions to correct any problems that may arise over time.

IT stakeholders will need:

y

The expertise to execute any coding required for data storage or for the rules used to move data, as in the case of data warehousing projects. Page | 6

y y y y y

An understanding of data models and database design (logical and physical design). The ability to abstract and generalize business requirements to fit data models and the political savvy to manage conflicting agendas when developing and refining data models. The ability to profile data errors at the technical level and define rules for cleaning data. An understanding of the data needs of the various business units. An understanding of the data requirements and limitations of enterprise applications.

Organisation
We recommend that there is a three tiered governance structure. The top tier will be formed by executive sponsorship layer and will consist of the overall programme sponsor (Exco member) and executive responsible for different data models e.g. (Global HR Director for employee data). These are essentially people who are sponsoring the programme, key decision makers and have responsibility of financing the initiatives. The next level is data governance committee consisting of Subject Matter Experts responsible for liaising between the committee and business units who have a vested interest in master data. The final level being the data quality team who are empowered by the data governance team to implement data quality project on the ground. They are also there to monitor quality and make recommendations to the governance team around policy and process. An example governance model for Employee Data Integration could be as follows:

Figure 2An example governance model for EDI

Each phase (EDI, CDI etc) will consist of different representatives in the data governance and data quality team. The only roles which remain across all phases are Executive Sponsor, CIO, Project Manager and IT Representatives. We also recommend a dedicated project manager be assigned who is responsible for theprocess. This individual provides a central point of contact for the data integrity team but also for the steering committee. Having a single person liaise between each of these two governance levels ensures that there is only one stream of communication flowing between the groups, to avoid the conflicting messages. In addition to this communicative role, the project manager also works to facilitate the progress of the project.

Team Responsibility
Page | 7

Governance Level The Executive Committee

Responsibilities y Consists of senior level executive(s) that are key stakeholders and final decision makers. y Sponsors and finances the governance program. y Establishes priorities and approves the data governance strategy. y Approves decisions and recommendations.
y y y

Data Governance Committee

y y y y

Consists of subject matter experts for each data set and from the IT department. Must have sufficient decision making authority. Decide what is to be done regarding data policy, procedure and standards development and implementation. Since these individuals are high-level planners that understand the business, they need to have the most input in policy development. Manages the data integrity team and reports to the executive leadership. Empowers data integrity team to implement governance recommendations. Resolves disagreements. Monitors costs and progress to keep project on track. Consists of data stewards/content managers that represent the needs of each business unit and IT representatives (e.g. data modeller; database architect etc.). Liaison between governance committee and business units. Works with the governance committee to help develop policies, data definitions, define business rules etc. Offers recommendations for governance committee approval. Implement data quality project decisions at the ground level. Monitor/manage data access and quality to ensure compliance with policies, procedures and standards. Train end users on policies and procedures.

The Data Quality Team

y y y y y y

The Project Manager responsibilities are to:

y y y y y y

Facilitate processes to resolve issues. Facilitate the inclusion of business perspectives in the process. Facilitate empowerment of stewards. Communicate recommendations. Communicate progress and benefits to operations. Communicate involvement in business planning.

Deliverables
The data governance team will be responsible for delivering and policing the following policies:

Page | 8

1. 2. 3.

Data ownership policy. This describes the roles and responsibilities of the individuals charged with the stewardship of information Data integrity and integration policy. This defines the standards of data quality and identifies the key elements that will be integrated and how this will be done Data security policy. Specifies security standards for the storage of data and decisions regarding access and usage. This should be a part of the information security policy.

They will also be responsible to drawing up a Master Data Management Map(for each entity) consisting of: 1. Definition of data sources and data quality problems 2. Constructing agreed data definitions or a shared data vocabulary(SDV) 3. Creating data dictionaries 4. Deciding on data entry standards and resolving disputes between business units that record and access the same data 5. Creating data quality standards & metrics for auditing 6. Decisions about meta-data management

Technology Architecture
Overview
In the previous sections we have generally referred to the governance around data with particular focus on people, processes and ?. Underpinning this governance and data quality programme has to be a robust, scalable technology architecture framework. The architecture must be able to support the life cycle of master data from inception/creation through to disposal and archiving. It must also allow for appropriate access to this master data for all relevant applications. In <Company Name>our IT infrastructure has evolved organically with the growth of the business and in a disjointed and piecemeal fashion. Applications are implemented to support a particular business need creating loosely integrated silos of information and adding to the infrastructure complexity. This can clearly be seen in our organisation that has over 10 different applications holding employee data. At an informational level this means conflicting versions of the truth and this has the knock on effect to the business which relies on this corporate information to make better decisions.

Identifying Systems
To understand what approach to take in building our MDM architecture framework we have to clearly understand the different kinds of systems in our environment. In pure MDM terms these are divided into three: 1. 2. A System of Entry (SOE): this is a system that creates master data for example dynamics is used as the system of entry for vendor data. A System of Record (SOR): this is the system designated as the authoritative data source and is responsible for updating master data. A true system of record will also handle data authoring. The SOR and SOE are in most cases the same system e.g CRM A System of Reference (SORf): is a replica of master data held synchronised with the SOR. It is only used as a read only source for data consumers.

3.

In <Company Name>the SOE and SOR are essentially the same application. For example Dynamics is both an SOE and SOR for vendor information; CRM is both an SOE and SOR for Customer information and Dynamics is a SORf for customer information; There is no one SOE and SOR for Employee data but this should reside with Snowdrop (or any new global HR system implemented), AD and iView should be SORf for employee data. In the data integration section above we outlined the best approach to architecting a solution in our environment. To reiterate making significant changes to our applications to read from a centralised data

Page | 9

store will not only be costly but will elongate the time taken to deliver successful MDM. In light of this we
propose what is known as a coexistence or synchronised architecture.

Coexistence Architecture
This architecture uses enterprise application (e.g. CRM, Dynamics), to synchronise application specific master data sets with one another in real time, while remaining in their native storage location. This creates a logical enterprise master data source, rather than a physical one. So to be clear, master data will be created and stored in existing systems while also producing an authoritative record of master data that resides in the MDM system itself. The MDM system in this architecture functions as a system of reference (SORf), as opposed to a system of record. The reason for this is that data is created and modified in multiple locations, unlike a true system of record. The MDM system record is created using scheduled batch update and/or web service updates. An MDM system s record of master data is synchronized with the master data across source systems in order to reconcile inconsistencies and guarantee that the data is current. The MDM hub also contains services to manage data conflict, reconciliation, and merging. Queries and updates can be performed within the MDM system itself. When the data is updated, synchronization processes allow changes to be fed back to the source systems ensuring consistency. Data can also be sent to downstream reporting and analytical systems and also to other system of reference applications e.g. AD or MOSS (see Figure 3). Coexistence implementations are recommended in IT environments like ours with loose and distributed infrastructures where the MDM system can manage master data and act as an authoritative source.

Figure 3 Coexistence MDM architecture

In the first instance we would recommend building and hosting the MDM on premise in one of our data centres. However there is no reason why this hub should not be hosted off premise in the cloud. Initial investigation has shown a number of organisations slowly adopting the concept of Virtual Master Data Management (VMDM) using current data virtualisation technology to build MDM hubs in the cloud.

Page | 10

Appendix A
Assumptions
Data quality
y
There has to be sufficient knowledge and political power to make decisions. The data quality team provides all the heavy lifting for a data quality project. To get things done they need to know what they are doing and have the active support of the executive. The data quality team needs strong representation by both IT and the business units that will be affected. Data quality management requires both technological solutions and changes in business processes. Having representatives from both groups of individuals will ensure that actions taken by the data quality team address the needs of all stakeholders. Policies and procedures are enforceable.There should be consequences for non-adherence to policy

Page | 11