James Thew - Fotolia
Looking for something else?
The first rule of data governance is you do not talk about data governance.
That's one of the tenets Anne Marie Smith sticks to when she works to break through companies' initial resistance to data governance and other equally stifling terms.
Smith, vice president of education and chief methodologist at data management consultancy EWSolutions, has spent her entire career in data management. She's contributed to a number of data management industry initiatives, including The DAMA Guide to the Data Management Body of Knowledge, and, over the past 20 years, has helped countless clients implement data governance plans. At 4 feet 11 inches tall, Smith jokes that she's more than knee-deep in data governance; she's "shoulder-deep in it."
Smith recently spoke with SearchDataManagement about when to implement a data governance plan, the problems companies experience when they skip this critical component and how to make data governance work to your advantage.
Some say big data lends itself to exploration and discovery more than to traditional business intelligence analysis. That means the data is modeled by data analysts before it is governed -- which seems entirely backwards. Is this a cycle that you typically see, and how do you help companies with that?
Anne Marie Smith: Yes, it is backwards. Data governance is about 85% reactive, and only about 15% is proactive.
We have many clients who ask us to help them do data governance after they've already initiated big data programs. They say 'we tried big data and it didn't work.' They thought creating two or three policies around data sharing was enough. Then we go in and discover all they have is data sludge … and we figure out why they can't get any actionable results.
What are some common challenges you see from companies who haven't implemented data governance?
Smith: Fear of the data police is one. People think, 'I don't want to turn into the data police if I lead this project,' or 'I don't want to be associated with it.'
Also, organizations that never had data policies don't want to have to start following them. Lack of executive support is another challenge; senior managers have to commit to a permanent data governance plan.
How do you balance giving people all the data they want without looking like the data police standing in their way?
Smith: We generally don't talk about practices, policies, standards and procedures because those terms engender the idea of the data police. Organizations that approach data governance using those terms are being counterproductive.
Instead, we talk to clients about how to use their data more effectively. We talk about removing the data barriers raised over the years, such as people not understanding how data flows through the entire organization, or the barriers that exist because of rules in separate parts of the organization. We say, 'Let's get together and discuss breaking down barriers.'
What we are doing is creating a new policy, and we are creating standards and practices, but we don't use those words. We talk about consensus. By changing the language away from restriction, you make [data governance] an inclusive process.
What are the biggest advantages of committing to data governance?
Smith: Data is as important an asset as money today. Some companies are selling data -- it's an asset -- and they need to govern that.
In general, when you have a good data governance program, an organization understands its data; where it exists, how it exists, how it is used, who is responsible for it, where it is going and what its value is. And not just in dollars and cents, in terms of usability and reusability.
With data governance, you have all of this information. Everyone knows the value of the data they control, they know what it means, where it came from and where it's going. The company can bank on the completeness and accuracy of its data. Without all of this information, [business intelligence] reports aren't trustworthy.
I imagine lack of data governance is a big problem for companies that use business intelligence tools, especially black box self-service BI tools where they don't necessarily know the methodology.
Smith: We have a lot of clients attempting to use self-service BI, and I see a lot of them stumbling as if they are in a dark room with one 10-watt light.
The structure of the data may be a problem, or the data is not well architected, or there is very little business metadata available. Combine those challenges with the lack of governed, managed data, and self-service BI becomes very difficult to implement.
Most people expect BI tools to be what they see in movies; they can point here and there and, all of a sudden, colorful results appear. The only place it works like that is in the movies.
Can you share an example of a case where a company did not implement data governance, and the problems that caused?
Smith: We had a client, an insurance company, that was trying to manage its offerings for auto and homeowners insurance, which were both offered online to the general public. We were asked to assess their metadata, data management, data warehousing, [etc.]
We made recommendations, including developing a data governance program for both of those lines of business. Those businesses were marketed to the same clients, so the customer information should have been shared throughout the organization. They didn't take [most of] our recommendations.
Four years after the assessment, they were acquired for their book of business, and they shut down. If they would have taken our suggestions, which would not have been costly, and would have been manageable with their existing staff -- plus one or two additions, in some cases -- their business would still be here, and it would have thrived.
Anne Marie, it sounds like your job can be really frustrating.
Smith: [Laughs] No -- when clients take our recommendations, it works.
We had a client in healthcare [the Mayo Clinic] that worked with us on data warehousing, metadata management and data governance. They combined data from multiple research sites and patient data from three hospital sites. [By combining the data], they discovered that patients with certain forms of cancer responded better to certain treatments. Doing this work actually saved lives.
That's quite a feather in the data governance cap.
Smith: It is.
What are some data governance best practice steps you can share?
Smith: Do an assessment; it doesn't take long. Most companies don't want to do an assessment because they are afraid. Nothing will bite you. I promise.
Don't be afraid to have someone outside the company look objectively. Most good data governance assessments, even for large companies, aren't laborious, and they don't take that long. I did one for a fortune 100 company that took 12 weeks [in all, including delivering the reports]. For smaller organizations, it might be four to six weeks.
Many organizations try to go it alone. They think, it can't be that hard. But there are lots of wrong ways to do data governance, and doing it wrong means you've spent a lot of time, money and effort setting policies that have to be undone eventually.
Data governance tools are only one part of an effective plan
How to build a data governance framework
Choosing the right data governance tools