by Sunil Soares
I work with a number of clients across multiple industries and geographies on their data governance programs. While every organization has its own challenges around data governance, the universal question is, “Where do we start?”
In this blog, I will discuss my industry-oriented approach to data governance that starts with a small set of attributes. The following ten steps use banking as an example, but they really apply to any industry:
1. Define business problem
Start by defining the problem that creates a business imperative for data governance. The business problem should be stated in terms that are highly relevant to your industry, your organization, and the lines of business or functions that sponsor your program. As an example, customer centricity might drive data governance at a bank that wants to find new customers and increase the number of products per customer.
2. Prioritize data domains
At the end of the day, data governance is about the “data” itself. It is sometimes easy to lose sight of that fact when you have to deal with data governance councils, charters, and organizational politics. A mature enterprise data governance program in a bank would want to address customer, product, financial, risk, and other data domains. Following through on the example from above, the bank should start by focusing only on customer data to support the customer centricity program. It might even want to focus on retail versus corporate banking customer data.
3. Identify critical data elements
Even retail customer data might be too broad because it may contain hundreds of attributes. In this case, the data governance program may identify a small subset of attributes such as date of birth, phone number, and address that have an outsized impact on business results at the bank. We refer to these attributes as critical data elements.
4. Quantify the financial benefits
This is the hard part but still very important. The data governance program should work with business stakeholders to quantify the anticipated financial benefits from data governance. This is the only way to get the business to assign data stewards and to take ownership of the data. In one instance, a bank found that it was losing millions of dollars in lost cross-sell opportunities due to inaccurate dates of birth. The bank validated dates of birth on a monthly basis with the tax department. Data quality processes were triggered when a customer with an incorrect date of birth applied for a new product. In these situations, the regulators insisted that the bank require its customers to physically visit a branch to produce the appropriate documentary evidence and fix their dates of birth. Needless to say, approximately 50 percent of these customers failed to visit the branch and the bank lost valuable cross-sell opportunities.
5. Create a RACI matrix
The next step is to build a RACI matrix that defines the Responsible, Accountable, Consulted, and Informed roles for data governance. In the above example, the bank created a Customer Information Management department that had overall accountability for customer data including date of birth. However, Branch Operations was the responsible party because it had the ability to actually fix inaccurate dates of birth. Compliance and Marketing were both consulted parties because they needed to ensure that the bank was not sending marketing mailings to minors. In addition, the Internet Banking group was also a consulted party because minors were not eligible for online banking.
6. Define the data governance organization
The RACI matrix drives the membership in the data governance council. It also drives agreement on the reporting structure for the data stewards. In the banking example, Customer Information Management, Marketing, Compliance, Internet Banking, and Branch Operations were all represented on the data governance council. In addition, the data stewards reported into Customer Information Management.
7. Write a data governance charter
The data governance charter acts as the constitution of the program. It should lay out the business problem, organization, and roles and responsibilities.
8. Establish a data quality scorecard with a handful of key metrics
The data quality scorecard should start with a handful of metrics based on the critical data elements. In the banking example, the data governance program produced a scorecard showing the number of inaccurate dates of birth on a monthly basis. The data quality scorecard was circulated on a monthly basis to the data governance council and the data stewards.
9. Stand-up a business glossary starting with a small number of key business terms
As with everything else, the business glossary should start with a handful of key business terms. Maybe just 15 or 20 business terms to start with. It can grow from there. For example, the bank focused on the definition of “customer” as one of the first terms in its business glossary.
10. Align with the technical architecture
Software tools can support and accelerate the data governance process. In the banking example, the data governance program worked closely with IT on various technical initiatives including master data management, data quality, and metadata.
Once you have cycled through these steps, you should circle back and expand the scope to focus on additional data domains and attributes. You can learn more about the banking example and the RACI matrix in this article, which is a modified extract from my book Selling Information Governance to the Business: Best Practices by Industry and Job Function (MC Press, 2011).