What is data governance, how can it be implemented, and what are the benefits of strong data governance? Our data management experts provide deep insights and answers to frequently asked questions about data governance roles, mandates and responsibilities.
In general, governance refers to the way the organization goes about ensuring that strategies are set, monitored, and achieved (Rau, 2004, p.35). Thus, governance implements a strategy by means of oversight and control mechanisms (Tiwana & Kim, 2015) and complements strategic as well as operational tasks: Strategy is doing the right things, operations are doing things right, and governance is ensuring that the right things are done right. In analogy, data governance links strategic data monetization initiatives with their daily execution to guarantee a return from data investments.
Usually, the following situations indicate that an organization suffers from unclear or non-existent responsibilities related to their data, meaning a lack of an effective data governance approach:
The problem with the term "governance" is that asking three different people what the word means, there will be at least four answers. The following definition reflects the real-world experiences from many companies and data initiatives, and is grounded in CC CDQ data management research:
The CC CDQ defines data governance as follows:
Data governance defines roles and assigns responsibilities for decision areas to these roles. It establishes organization-wide guidelines and standards, and it assures compliance with corporate strategy and laws governing data. (Weber et al 2009)
Best Practices: After defining responsibilities on paper, select and train the people who will take care of the data management activities on a day-to-day basis. It is recommended to select these individuals with utmost care. All too often, the people already overloaded with work also end up taking care of data. If that is the case, it will be impossible to achieve the agreed objectives, for instance, related to data quality.
The benefits of an effective data governance framework are threefold:
Data leaders face several challenges when setting up data governance. One of them is that there are only a few guidelines on implementing data governance and that knowledge remains mostly tacit. Here is the good news: Defining a data governance framework isn’t rocket science. CC CDQ's reference model for data governance can help in developing and implementing a data governance framework. A proven approach to follow: Firstly, defining which data management activities are to be undertaken by whom in the organization. Secondly, deciding how many people are needed to run the governance plan on a day-to-day basis. Thirdly, setting the mandate and decision-making rights for the people defining and monitoring the governance policies and standards. Fourthly, defining an implementation plan, including training, communication and change management activities. Finally, select the people executing data governance.
A data governance program must follow a clear goal. Hence, it is of utmost importance to clarify the deliverables in the beginning. The deliverables concern different data management related decision areas in the enterprise, most importantly the data strategy, the data management portfolio and roadmap, the data governance framework, data models and architecture, and the data lifecycles. Once the deliverables are set, the relevant data management activities must be defined.
The required activities are defined in CDQ's data management process map. The map includes operational activities (e.g., creating and maintaining data), tactical activities (e.g., building up and running data quality management) and strategic activities (e.g., developing a data management strategy). To be effective, the activities from the process map must be assigned to employees in the organization.
We do not recommend assigning the activities directly to individual employees. Instead, we advise to allocate certain sets of activities to specific data management roles (e.g., a Data Steward) and to assign these roles to individual employees. A proven tool that helps to assign the responsibilities to Data Governance Roles is the RACI matrix (i.e., Responsible, Accountable, Consulted, Informed).
To obtain management approval and budget, it is necessary to estimate how many employees are needed to execute effective data governance. The best way to determine that figure is to estimate the overall workload. In a typical data governance setup, there are some full-time positions, typically in a central data management team. In addition, there will also be some employees in the business units and support functions that cooperate with the central team. However, these are typically not full-time data management positions.
To be effective, the people executing the data management activities need a clear mandate. The mandate defines the authority of the data management roles. If it is left unclear what they can or cannot decide, the entire governance setup will not work. In addition, it is necessary to define decision-making boards where stakeholders of the organization come together and align on data management activities and issues. This is needed because data is a shared resource, that is used by various organizational units.
Implementing data governance requires an organization to change. Therefore, change management activities are required to support the implementation. These include an effective communication strategy tailored at different stakeholders. In addition, a plan to train the employees involved in data governance is also needed.
The best data governance concept is likely to fail if the wrong employees for the data management roles are selected. Business understanding is key and often more important than technical/IT knowledge. At the end of the day, data managers often bring people from diverse backgrounds together, so strong communication skills are a must.
Two important design principles are fundamental for any data governance design:
*Data definition includes business and quality rules, data access policies, data lifecycle and the conceptual data model. From the data point of view, it thus provides the input for the authorization concept, the risk management and compliance.
Data governance introduces new roles (or makes them explicit). The following role model is grounded in CC CDQ research and based on practical experiences in many companies and data initiatives:
|Chief Data Officer||
|Head of Data Management||
|Data Definition Owner||
Ensures that business requirements are fulfilled, and data are compliantly accessed and used:
|Data Content Owner||
(Business Data Steward)
(Technical data steward)
|Data Quality Manager||
|Data Documentation Manager||
Source: Based on Fadler & Legner (2020)
Assigning these data governance roles in an organization depends on many factors – most importantly, the maturity of the company and the mandate for data management. In practice, many variants can be found. But independently on how roles are assigned, three groups must very closely work together: data management teams, business functions/lines as well as the IT organization.
With the increasing recognition of data as a strategic asset, the role of the Chief Data Officer (CDO) - also called head of data and analytics or chief data and analytics officer – is becoming of major importance in enterprises. This role creates the bridge between data management and analytics-oriented teams, while steering the overarching data strategy at an enterprise level and fostering alignment with business and IT stakeholders.
In case a CDO does not exist, typically the Head of Data Management can be found in enterprises who steers the central data management team, and is responsible for the data management strategy, portfolio & roadmap and accountable for their implementation (data governance framework and high-level data architecture).
A central data management team typically consists of several Data Stewards. The Data Steward is responsible for specific areas of responsibility’s data definition. The data definition includes business and quality rules, data access policies, data lifecycle and the conceptual data model. From the data point of view, it thus provides the input for the authorization concept, the risk management and compliance. The Data Steward is complemented with Data Architects, Data Quality Managers, and Data Documentation Managers.
The Data Architect supports the data steward by designing, creating, deploying, and managing conceptual and logical data models, as well as with mapping to physical data models. The Data Quality Manager is responsible for data quality metrics and methods and acts as a subject matter expert across data domains.
The Data Documentation Manager is responsible for documenting data and data related standards and frameworks. This role becomes an essential driver of data catalog initiatives, for instance.
To be effective, any central data management team needs to be linked to the business units and support functions. After all, this is where most of the data is created and used. Most importantly an effective data governance design requires data ownership to remain with the business functions/departments. Data ownership clarifies the fundamental accountabilities for data in the organization. Two types of ownership can be distinguished here - the data definition owner and the data content owner. However, depending on the size of the company
Depending on the size of the company, it might be necessary to create one or more roles to ensure the link. Some organizations have established a Business Unit or Divisional Data Steward who ensure that the activities of the central team are aligned with the needs and activities of the business units and support functions. Other companies have defined Data Experts who are contact persons for the central team because they have very specific knowledge about certain data attributes (e.g. health and safety data, tariff numbers for customs). In most organizations, data is created and maintained in the business units and support functions by Data Editors under supervision of Data Content Owners.
Since data is processed and stored in IT systems, the link between the data management roles and the IT roles is important. Therefore, it is important to align the Data Steward roles with the IT roles, for example a Demand Officer or Solution Architect. Data Management roles also need to closely work with process management roles, for example a Process Owner and Process Manager. After all, business processes require high-quality data to operate efficiently. In addition, process changes typically have implications on the data that is used in the processes. Finally, with analytics becoming ever more important in most organizations, the data management roles must be linked to the Data Scientists and Data Analysts roles.
Data is a shared resource, that is used by several business units and support functions. Consequently, data-related priorities, requirements and issues must be aligned between them. This alignment happens in data governance boards. Most organizations require one board for strategic data management topics and one or more boards for tactical data management issues.
This board (the name varies across organizations) is the central decision-making board in data governance that aligns and approves the data management strategy, annual objectives and budget. It also tracks the progress of the data management activities and resolves issues that had to be escalated. Oftentimes, the board consists of data definition owners from the business units and managers of the support functions as well as the Head of Data Management / Chief Data Officer. It is possible that these roles already meet in other boards or committees. If so, it should be considered whether an additional strategic data management board is really needed or whether the topics to be discussed can be integrated into the other boards or committees.
Data governance council(s) are needed to align more operative data management issues. These can include changes to the data model that affect several business units or the prioritization of data quality issues that must be resolved. Depending on the size of the firm, it might be necessary to have several councils. If that is the case, they are typically established per data domain (e.g. one council for customer data, one council for material/product data etc.) Data governance council(s) meet more frequently than the strategic board. Members are the Business Data Stewards, Data Quality Manager, Data Documentation Manager, and Data Architect from the central data management team as well as the Business Unit / Divisional Data Stewards.
The mandate defines the authority granted to data management roles within the organization, and from this, the responsibility. This is particularly critical for the central data management team. Are they allowed to enforce the standards and guidelines for data creation and maintenance or are they dependent on the goodwill of the business units and support functions? The type of mandate has a significant influence on the overall impact of data management on the organization. Like the mandate of a project management office (PMO), four types of mandates can be distinguished.
It is not uncommon that the authority of the (central) data management team is not defined. It is unclear what they can decide or how they can influence the organization to improve data management practices. If there is no clear mandate, discussions about data management priorities and initiatives tend to be endless. The problems with having no clear mandate for data management is often the trigger for defining and implementing a data governance.
If the mandate is to support, the data management team provides templates, best practices and access to information. However, the influence on the data management activities in the business units and support functions is very low. They get to choose if they want to cooperate with the (central) data management team or not. Since data is a shared resource that is used across organizational boundaries, the mandate to support is not very effective.
In this mandate, data management has the authority to align and coordinate all data management activities across the entire organization. This helps to reduce duplicate work. A central data management team with a mandate to coordinate can also advise on methods and best practices. However, it cannot enforce them.
In a directive mandate, data management has the authority to enforce data management standards through various means. Firstly, it defines the framework, methodologies and tools to be used for data management across the organization. Secondly, it also monitors whether the standards for creating and maintaining data are upheld. If not, it can sanction non-compliance, for example by revoking access to IT systems.
Best Practices: The mandate should be defined early in the data governance journey. If the difficult questions about authority and decision are not discussed upfront, they will return during the implementation of the data management initiatives. Since these discussions can’t be avoided, we recommend having them early in a data governance journey.
How and where is the required data quality defined? What are acceptable values for a specific data attribute and what aren’t? How should data be documented as metadata? Data Governance related decisions refer to some fundamental principles of data management. They include standards and guidelines, for instance related to data creation and maintenance, but also more general policies and principles that make employees understand the true importance and value of data.
Guidelines, policies and standards (sometimes also referred to as data management rule books) define how quality is measured, how data is documented (metadata) and how the data life cycle managed. But most importantly, they comprise documents that define, how to create and maintain data correctly. For each data domain, they describe the key data objects with permitted values for the data attributes. For example, if we have a German postal address, is it permitted to shorten the word “street” to “Str.” or is it mandatory to write “Strasse”? Similarly, should the legal form of the company be part of the name or not? These guidelines are relevant for the entire data life cycle and for all systems in which the data is stored or used. Good guidelines, policies or standards are the foundation for high-quality data that is also harmonized across the company. The key to success is to train the organization and to monitor whether users adhere to the rules. Some rules defined in the guidelines can probably also be coded as validation rules into the IT systems.
Effective and efficient data management requires organizations to change. As known, change is hard. Many change initiatives fail simply because employees don’t understand why they need to change. This is where data management principles come into play. They describe the data management aspiration in easy to understand words and give enough concrete guidance on how to deal with data on a day-to-day basis. Principles also help to communicate the abstract topic of data management in a more tangible way.
Best Practices: No one comes to work in the morning to create poor data quality on purpose, but data quality is often the result of ignorance or negligence. Guidelines, policies and standards define what good data quality looks like. Data management principles help to explain why data matters and why it is everyone’s responsibility.
Do you have any questions? Our data management specialists will be happy to help you with any questions you may have about data governance. Just contact us!