Keys to Data Governance Success

Data governance practices are essential for managing data as an asset. These practices establish repeatable, measurable business processes and manageable policies for improving data quality. Data governance helps companies meet regulatory compliance mandates while improving revenue opportunities and customer and partner relationships through the power of higher quality information.

Despite the universal acceptance that data governance practices are worthwhile, businesses have historically bypassed the discipline in favor of other initiatives that more directly impact the bottom line or provide an immediate return on investment (ROI).

The growing requirements for data privacy and security have altered this landscape for the modern enterprise. Today's commercial businesses are being driven into data governance action by legislation such as the Gramm-Leach-Bliley Act and the Sarbanes Oxley Act.

Even with these legislative motivators, data governance is at best inconsistent. Most businesses require a guiding hand to assist them with the best ways to achieve success. Once again, companies are discussing the value of data to corporate operations and to gain insight and visibility into corporate performance. However legislative motivators have created a tipping point where companies must ensure they are protecting consumer privacy preferences and other personal data.

Why Companies Struggle

Although some companies today are doing a good job with data governance, they are the exceptions and not the rule. The big question is why do companies continue to struggle with this practice when the proper management of company data will increase revenue, improve customer and partner relationships and company operations, and assist with regulatory compliance issues?

One reason is that many businesses are short sighted because they operate on a quarter-by-quarter revenue model. In addition, most publicly traded companies are focused on ensuring strong short term stock performance. These objectives conflict with data governance initiatives, which usually have a less direct ROI cycle, and are competing for precious resources with other corporate programs that promise more immediate revenue, efficiency, or profitability benefits.

Another reason is that data governance and sharing data between product groups or divisions of the same company can be highly political. Often, business leaders that become territorial about their customer data are unwilling to share data, even for the betterment of the greater good. For instance, imagine a software company that offers two types of consumer applications. There may be natural relationships between the products which could afford the company the opportunity to cross sell to an existing customer base, but the business executive of Division A refuses to share his customer data because he is not rewarded based on the performance of Division B. This type of "turf war" happens frequently and extinguishes an opportunity for the company, as a whole, to perform to its best advantage.

A third reason is that most businesses that have tried to implement a data governance practice in their organization have taken the wrong approach. Some companies have tried assigning the program to a single individual who ultimately fails because the job is too big and broad for one person. Still other organizations have knitted together a coalition of interested director-level parties, which has resulted in limited success because this group doesn't typically have the budget authority needed for a proper data governance program, nor the influence to shift corporate priorities. The most successful initiatives have taken a top-down strategy of appointing an executive sponsor whose "day job" it is to get their arms around data quality. But even this approach requires the coordinated efforts of IT, business management, finance and other functional units like marketing, product management and sales.

Lastly, and most importantly, many companies feel that, in order to achieve data governance success, they must take a "boil the ocean" approach. Trying to tackle all of their data governance issues in one swoop often becomes overwhelming, and requires a time investment of several years, with costs escalating before any true return is realized.

Getting it Together

An effective data governance program includes the people, processes and policies necessary to create a single, consistent view of an enterprise's data. These programs require the coordination of a myriad of people across organizational and political boundaries, all of whom have other "day jobs." Enlisting senior executives and other employees into a data governance program requires a corporate priority shift, usually away from more "gratifying" tactical issue resolution toward more nebulous and difficult to quantify data quality initiatives.

Instead of employing a top down, "all or nothing" approach, which requires a complete shift in corporate culture, businesses are better served with an iterative approach to data governance. They should choose a smaller data set, like customer or product data, and focus their efforts on investigating and fixing it, then determine what worked and establish policies that can be used to tackle another data set. This more cost effective iterative approach enables companies to secure immediate results within months instead of years, while putting into place the scaffolding needed to help manage data quality across all systems. An iterative approach also allows companies to keep an eye on the quarterly revenue ball while making gradual improvements that will ultimately have a favorable impact on the bottom line. Think of it as "Agile Data Governance."

Businesses that combine this "bite size" strategy with a strong team approach to data governance will be most successful. The most sensible approach is to appoint a senior executive as committee chair. Ideally this executive would report directly to the CEO and have the clout to dedicate budget, alter corporate priorities and eliminate "turf wars." The chair should oversee a board comprised of other executive leaders who have responsibility for each of the company's lines of business. In addition to this group, an effective data governance team should include individuals representing horizontal functions such as finance, human resources, IT, accounting and marketing, and a group of data experts including a data owner, data steward, data architect and data modeler, and a group of data analysts. These data experts have the following roles and duties:

  • Data owner - establishes policies and owns data quality for one or more master data domains, such as customer data, product data, portfolio data, location data, etc.
  • Data steward - implements and enforces policies and business rules, and corrects data quality problems including matching records, replacing bad data with good data and making "survivorship" decisions if more than one record for the same person exists.
  • Data architect - evaluates and modifies system components to alleviate data quality problems
  • Data modeler - captures and documents business rules that determine data quality
  • Data analysts - discover and research problems for the data owner(s) and investigate data quality on a record-by-record, value-by-value basis to look for exceptions, duplicates, etc.

A business can further its success by dovetailing an iterative approach to data governance with some basic committee guidelines. Team members should stay focused on the end result and avoid endless philosophical disagreements about the meaning of data versus information, the exact definition of an entity or attribute, or what type of representation to use (entity relationship diagrams or class diagrams). The program will be much more successful if the team can deliver something quickly that may not be 100 percent perfect in its first iteration. The goal of the team should be to complete each iteration rapidly, learn as much as possible, employ "just enough" process to plan and measure results, make adjustments for the next iteration, and improve the process with each iteration.

Defining Data Governance Processes

With an iterative approach and the right team in place, commercial businesses are ready to define their data governance processes. These processes typically begin with a business assessment phase and conclude with deployment and ongoing maintenance:

1. Business assessment - Articulates various "data problems" and the value of improving them. This step maps to high level business processes and helps companies prioritize the pain, issues, costs and value of data. Business assessment results in the definition of the biggest areas of pain, opportunity and risk, and quantifies the value of fixing problems.

2. Data architecture - Explains the "ecosystem" in which data are created, maintained, propagated and leveraged for business purposes, and produces a data dictionary and catalog. This step maps data sources and targets, transformation points, moves and transfers (messages, files, ETLs), and areas of use in business transactions or decision making. Data architecture results in the measurement of the "complexity" of the data landscape being examined, and influences the definition of the "optimal costs and benefits" of fixing data.

3. Proof of concept (if applicable) - Data governance teams move through the steps four through six to perform a quick data quality and remediation process, which includes acquisition, analysis and remediation.

4. Data acquisition - Analyzes the data available to select data for transformation. Data acquisition captures data from various sources and produces output for consolidation. At this point, it may be necessary to model data under examination, but enterprises should be careful not to model all of an organization's data. It is recommended that companies leverage industry standard models or schemas for ideas on how to represent common highly-shared data, and map all data taken from source systems to this standard for sharing. This approach keeps system-specific schemas private to systems, while promoting a standard for managing and moving data.

5. Data analysis - Profiles source data and models to identify inconsistent or unclear data for clarification and establish policies and business rules. Data analysis establishes algorithms for determining the uniqueness of data and defines data quality patterns and business rules for cleansing data. It also creates ongoing procedures for managing data quality such as triggers, tasks, stakeholders, workflows, processes, approvals, escalations, etc.

6. Data remediation - Repairs data according to specified business rules and data quality processes. Applies business rules test results to data, and publishes findings on data quality results, duplication rates, data error rates and patterns, volumes, and adjustments to data quality rules.

7. Evaluation and recommendations - Produces a final report that evaluates costs, benefits and scope of changes required to alleviate data quality problems. Determines the implementation phases, expected costs and results, and finalizes the commitments of the project team to make changes.

8. Deployment - Makes systemic, messaging, process and policy changes in participating systems that follow normal development-test-production lifecycles. Ensures that systems have migrated to new data values, while protecting data integrity.

9. Ongoing maintenance - Evaluates business rules and policies, captures and reports on metrics, and measures progress. Data governance team members may choose to make changes to policies, rules and workflow processes, or may decide to automate something that was once manual, leading back through steps one, two, five, seven and eight.

Closing Thoughts

Businesses that see the value of a data governance initiative, but are hesitant to begin the process because of concerns about costs or ROI cycles don't need to abandon their efforts. Instead, they should employ an iterative approach that keeps the scope of the project small and focused on one portion of business data, such as customer or product information. A focused approach enables an organization to realize success quickly, while establishing policies and creating frameworks that can be applied to other parts of the business in the future. Of course a successful data governance initiative also requires a strong team formed from various business leadership positions and strong data resources. Although an incremental approach still requires system changes and results in integration and architectural challenges, companies that choose this route have a much greater chance for success than those that try to "boil the ocean."

About the Author

Marty Moseley serves as chief technology officer at Initiate Systems where he is responsible for the company’s strategic technology direction, development and future product evolution. Initiate Systems, Inc. is the leading provider of customer-centric master data management software for companies and government agencies that want to create the most complete, real-time views of people, households and organizations from data dispersed across multiple application systems and databases. He can be reached at and additional information on Initiate Systems is available at

More by Marty Moseley