Without effective data governance and data management, big data can mean big problems for many organizations already struggling with more data than they can handle. That “lake” they are building can very easily become a “cesspool” without appropriate data management practices that are adapted to this new platform. The solution? Firms need to actively adapt their data governance and data management capabilities from implementing to ongoing maintenance. Here are 10 steps to follow on that data journey to smarter, faster business decisions and results that big data can empower.
1. Data Governance -- Inside and Out
Too often, data governance is primarily focused only on inbound data quality improvements, with no proactive stewardship, such as cataloging incoming data feeds, defining data standards and ensuring transparency of business rules to maintain lineage. The result can be that downstream analytic environments still receive inconsistent data.
Avoid this problem with good onboarding processes and follow up ensure data is properly cataloged and easily discoverable and establish simple easy to follow methods to ensure your users are part of the governance process and contribute their knowledge to the effort. Data governance is more than a small group of people tasked with making sure that the quality of data coming into the organization is good. It’s an enterprise effort where many can contribute.
2. Big Data Roles
Just as with all other data in your environment, you need a clear sense of who the key stakeholders/decision makers are. Roles may change as the data moves through your ecosystem and during its lifecycle, but they should be well understood, nonetheless. As you embark on your big data initiatives, identify these stakeholders as soon as you can, and be prepared to refine and iterate as you go.
3. Who Owns it?
Good data governance is predicated by good, business-driven data ownership. In practical reality, though, it can often be difficult to identify a single correct data owner. Let’s face it, data is being generated by individuals but owned by the firm. So, it’s common to want to start with what you know, which is often by identifying key stakeholders and stewards who are invested in the use of the data. But that can’t go on indefinitely.
Businesses need to actively managing the ownership issue to ensure that a committed business owner is soon identified. Organizations must establish timelines and regular check points, and begin to measure the area being governed with key milestones like assigning clear accountability to ensure progress is being made, and ensuring clear measurements are in place. Make it transparent. Keep senior management involved.
4. Flex Up for Big Results
Data governance and data management processes should adapt to support the needs of big data users and big data technology. They can be flexible and should take into consideration context and discovery in addition to the more rigid operational and transactional needs. Apply adaptive governance processes that can support various degrees of rigor and oversight to match.
Ensure your reference information architecture is updated to support big data concepts such as unstructured data streams.
Metadata management capabilities should be enhanced to include/correlate all of the basic metadata components as well as support rich data types in the form of tagging’.
Once you promote your solutions to production, you will want them consistently used, so it’s critical that you put a lot of effort into communications, education and training on the appropriate use of Big Data.
5. Include Unstructured Content Management
Unstructured data has always been a part of business operations, but now that we have better technologies to explore, analyze and infuse this content into business processes and insight work, it’s critical that we finally take the step to formally include its management. Most firms struggle here and the ability to mine this information has been difficult at best.
Don’t take the “easy route” and simply rely on big data technology to be your only formal data management process for unstructured data. Over time, as more and more unstructured data is added, being clear on what is good, what is bad, where it came from and the decisions it supports consistently become increasingly important.
6. Business Value View From the Middle
Governance is increasingly well received at senior management levels, but middle management is almost always skeptical. They have real deliverables, with hard dates to meet, and limited resources. They are often the cause of slow adoption and maturity. Return on investment is often hard to define for any governance activities at the enterprise.
The business value of data governance is different at this level. A good data governance implementation will identify key integration points with existing processes and strategic initiatives that have a well defined value proposition and help the organization understand how to leverage well managed data to their advantage. Use early adopters as data champions. Get the word out as to what works, and even what doesn’t work. Help make the business value clear.
7. Set the Dial on Precision
Whether it is a POC or a project that has moved into your mainstream business processes, be very clear on what you intend to do with the data and what level of quality and precision you desire. With big data, this can be a spectrum. You may have exploratory data that supports early decisions and production grade data that supports financial reporting Higher quality and precision requires stronger data management and oversight. As your organization matures with big data, consider establishing a quality- or precision-classification approach that will allow the data users to understand what they’re using and adjust their expectations accordingly.
8. Be Purely Practical
Too many struggles of implementing data governance stem from purism when the assembled team wants everything done by the book, but the organization lacks the skills, understanding, and business buy-in to support their demands. These purists also often want a level of detail that requires major resources to comply; projects and operational areas just cannot absorb these additions in one fell swoop and funding often becomes a stumbling block.
So what can organizations do? Take a practical approach. Use an onboarding process that teaches the simple skills necessary to do the job. Help the business and I/T resources learn how to make governance work as a natural extension of their current work. Prioritize what needs to be done and at what pace. Make business and I/T drive the workload and timing. Taking a change management approach to data governance will help you calibrate your program to the pace your organization can handle it while showing gradual improvements that build momentum.
9. Formalize the Hand-Off
If you’re doing one-time analysis or a full-fledged throw-away pilot, this doesn’t apply, but for most organizations, their initial big data work quickly evolves, and they find a need to sustainably leverage the highly valuable information they’ve mined. This means taking the work done in your sandbox environments and formalizing it to run in your production environment. Establish an agile method for promoting’ your data development efforts to usable decision-support. The more you do it, the more demand will increase.
10. Overseeing Evolves
What needs to be overseen changes as the organization matures. For instance, early on it’s easy to measure simple onboarding. Over time, the need to add Meta data and data quality metrics will arise. Companies will also look to assess data impacts in project oversight and in technology activities. To evolve with these needs, organizations need to be proactive, have a small team either dedicated or dynamic depending on scale and capacity that gets together regularly to assess the company’s progress and identify ways to move the ball forward. Develop a simple set scorecard that is easy to report on to senior leadership. They like tangible metrics, especially if they trend upwards, and so should you.
Avi Kalderon is NewVantage Partners’ practice leader for Big Data and Analytics and heads its New York Office.
This blog entry has been reprinted with permission.
Readers are encouraged to respond using the “Add Your Comments” box below.
The opinions posted in this blog do not necessarily reflect those of Insurance Networking News or SourceMedia.
