The Wikipedia article on data governance gives a good definition and an overview of the related topics. But although you
may find a few hints on how data governance impacts the business intelligence
and analytics practice, the article is living proof that the link data governance
with BI and Analytics is not really on the agenda of many organisations.
Sure, DAMA
and the likes are reserving space in their Body of Knowledge for governance but
it remains on the operational level and data governance for analytics is considered a
derived result from data governance
for on line transaction processing (OLTP). I submit to you that it should be the
other way around. Data governance should start from a clear vision on what data
with which degree of consistency, accuracy and general quality measures to
support the quality of the decision making process is needed. In a second
iteration this vision should be translated into a governance process on the
source data in the OLTP systems. Once this vision is in place, the lineage from
source to target becomes transparent, trustworthy and managed for changes. Now
the derived result is compliance with data protection, data security and
auditability to comply with legislation like Sarbanes Oxley or the imminent EU
directives on data privacy.
Two observations to make my point
Depending
on the source, between 30 and 80 percent of all Business Intelligence projects
fail. The reasons for this failure are manifold: setting expectations too high
may be a cause but the root cause that emerges after thorough research is a
distrust in the data itself or in the way data are presented in context and defined
in their usability for the decision maker. Take the simple example of the
object “Customer”. If marketing and finance do not use the same perspective on this object,
conflicts are not far away. If finance considers anyone who has received an
invoice in the past ten years as a customer, marketing may have an issue with
that if 90 % of all customers renew their subscription or reorder books within
18 months. Only clear data governance
rules supported by a data architecture that facilitates both views on the object
“Customer” will avoid conflicts.
Another approach:
only 15 – 25 % of decision making is based on BI deliverables. On the plus side
it may mean that 75 % of decision making is focused on managing uncertainty or nonsystematic
risk which can be fine. But often it is rather the opposite: the organisation
lacks scenario based decision making to deal with uncertainty and uses “gut feeling”
and “experience” to take decisions that could have been fact based, if the
facts were made available in a trusted setting.
Let’s spread the awareness for data governance in BI
Many thanks
in advance!
