If you hear someone saying “D&A” it is not “DNA.” D&A stands for “Data and Analytics.” (Not “Depreciation and Amortization” or “Drug and Alcohol.”)
The term “Information” has been confusing for years, it led to many information-interpretation debates: what’s the difference between data and information? is information about analytics? how do we manage information if we don’t agree what it is? “Data and Analytics” is an unambiguous term.
With the D&A’s rising ubiquity, there are also rising concerns about balancing data and analytics governance. I recently moderated the roundtable on this subject. Below are the topics discussed at this roundtable:
- Reestablishing data governance
- Expanding data governance to analytics governance
- Starting information governance
- Governance of shared data
- Bimodal governance with focus on Hadoop
- Decentralized D&A governance
- Including analytics in purview of the Data Governance Council
- Engagement of data stewards
- A governance evolution: from compliance to value
- Consolidation of data governance.
The subject of data and analytics governance is new to most organizations, and there are more questions than answers. In Solution Path: Implementing Big Data for Analytics, I wrote:
Data governance expands to analytics. Analytics governance is a positive force that defines the rules, standards and guidelines for the enterprisewide use of analytics.
Of course, data governance is a huge part of analytics governance, because your analysis is only as good as your underlying data. But there are also specific issues around analytics. Examples include (but not limited to):
- Developing trust principles and standards for both big data and based on the data analysis (believe me, some people blindly trust the models)
- Probabilistic thresholds for models validity (and mindfully adjusting the eternal tension between false positives and false negatives)
- The number of iterations beyond which the law of diminishing returns kicks in, while teams still get carried away by getting closer and closer to… what? (the right answer, time limits, budget limits, patience limits, all of the above and beyond)
- Models’ versioning and change management (critical to reconstruct specific point-in-time outcomes depending on data dynamics and on the iterative model version)
- Monitoring models and analytics outcomes (models tend to degrade over time, as I described in Big Data Analytics Failures and How to Prevent Them), in the “Question Your Data” section, e.g. analytics and data are always tied)
- Expansion of analytics to new users (who do not have the slightest idea of data quality at best or regulatory compliance at worst)
- Policies for combining data sources: when one source is compliant, but together with two others it reveals sensitive information.
- Escalation procedures in case something unexpected happens (for example, an automated decision might make a bad effect on your company’s reputation)
The list of concerns and actions goes on and on. But the principles of governance centered on people remain the same, because, above all, governance is a discipline that depends on people. Through Gartner’s own data and analytics, I learned that my recent document EIM 1.0: Setting Up Enterprise Information Management and Governance turned out to be more popular than I expected. And I know why. Because EIM and governance are going from ivory towers of boredom back to earth full of excitement brought by D&A. All of a sudden, everybody wants governance! Governance is what makes D&A part of the company’s DNA.
Follow Svetlana on Twitter @Sve_Sic
The post Balancing Data and Analytics Governance appeared first on Svetlana Sicular.
