So you’re ready to create or maybe even revise your credit union’s data governance program. You have already proven that data is a valuable asset to your business and your members. Now is the time to layer on additional processes and procedures to create structure for your data management and analytics practice.
While this guide isn’t meant to be an end-to-end blueprint for starting from scratch, it will give you essential pointers on organizing and managing your data warehouse. Ultimately, your data governance efforts should build trust in your business intelligence efforts, which helps you make decisions based on historical data trends.
Incorporate these five data governance tips to get started:
1. Data quality rules
There’s a reason data quality comes in at #1 on this list. Data quality functions as a programmatic basis for writing, applying, and assigning accountability for a set of business-defined “rules.” We recommend automating and tracking your data quality – typically this is deployed alongside your data warehouse or data lakehouse. It uses metadata and timestamps to track your progress over time so you can measure data flaws against successful data cleanups.
An example of a useful data quality rule would be “open accounts that are missing an email address”. Your system is designed to allow new accounts without email addresses, but perhaps your marketing department wants to reduce the number of accounts with missing email addresses so they can better communicate with their members.
Marketing suggests running a direct mail campaign encouraging members to link their email address to their account to take advantage of an exclusive offer. Let’s say 15% of recipients respond to the campaign and take the desired action. Now we can both measure the success of the campaign and track the improvement of this specific data quality rule.
2. Clear definitions of each source
Mapping every field and table from every data source in your data model is another essential part of the data warehouse process. Documenting them will help you in the future when new fields are added or removed from the source or target systems. All in all, it’s another way to document your business logic in written form.
Many times the mapping can be clear and obvious, but other times not so much. If multiple source systems all have a unique “Member ID” field for each member record, you may need to choose one as the global primary key and store the others as associations. When really getting into the weeds, you’ll also need to consider destination and target data types. For example, if your “Member ID” field is alphanumeric in the source system, you will need to ensure that the target data model has a field that can contain alphanumeric characters and not just integers.
This process is a collaboration between business users who understand the business logic and all the important calculations alongside an experienced data or business analyst who can translate everything into SQL or other common database languages.
3. Data Dictionary
As an additional source for your definitions, your data dictionary will be a great asset for documenting specific terms, calculated fields, and other important data resources. Compiling a data dictionary is important, especially as your internal user base has more access to dashboards and reports. Supporting this deployment to more business users with a data dictionary will save your team a lot of time and secure your investment against turnover or loss of knowledge over time.
An example in banking can be taken from the Federal Reserve Bank of New York. Take a look at the data dictionary he published with an Equifax credit report to provide additional context for anyone consuming the information in the report itself.
Building a data dictionary from scratch is quite a big undertaking for any business. One way for credit unions to avoid tackling this problem alone is to leverage a CUSO with a credit union-specific data model that contains many shared definitions. From there, the company can define custom fields or calculations in addition to what is shared within the industry itself.
4. Referenceable hierarchies
An important aspect of data governance is managing and accounting for who has access to what. At the same time, you will want to make sure that the people who you need access to have it. One way to achieve this is to use specific reference tables.
Suppose the loans department has a long list of different types of loans that it uses on a regular basis. The loans department also strives to introduce new products to the market, so it must be able to add new types of loans on a semi-frequent basis. One way to grant them this access is to develop a process for them to update and upload, for example, an Excel file to a secure location that updates a reference table with the latest types of loans.
The loans department does not need administrative privileges to update the database itself, but instead has a predefined process to update a specific reference table with a spreadsheet easy to use. They simply add the new loan type code and loan type description, upload the file, then the database checks the file and updates the reference table without any further intervention from the analytics team.
This approach combines the secure and controlled access requirements of the data governance program, but still gives the loan department the flexibility it needs to perform its day-to-day roles and tasks.
5. Adoption Dashboard
How are your users interacting with the dashboard you created? Who are your top five experienced users? What are the most (and least) used dashboards?
These kinds of questions will begin to arise as your data analysis program matures. You’ll want to know how your internal users engage and engage with content they’ve requested in the past.
Creating a user adoption dashboard is one way to track usage trends over time when investing heavily in your data analytics. One way to use it is to manage paid licenses. If a user no longer logs in, you can easily review the data and swap a license with another executive or manager who will be using it more actively.
Returns on investment
Investing time and energy in your data governance program pays long-term dividends. You will better understand trends in historical data and increase your ability to act on insights discovered from focused data analysis. Review these tips at your next data governance committee meeting and take control of your data journey.
Judith Bernholc is the Customer Success Manager at Arkatechture, a consulting and data services company based in Portland, Maine.