Database Reference
In-Depth Information
•
Informationiscurrent
Only the most recent and relevant data is
stored, and old data is either archived or deleted. Big Data often
presents new challenges for lifecycle management, as Big Data is often
very time sensitive and loses its value quickly; it may be the case that
the lifecycle of the entire group of Big Data (such as social media data)
isn't actively governed, the entire dataset is analyzed, and then deleted
once the analysis is complete. That said, there are also considerations
in a Big Data world where you want to store a corpus of data to build
better predictive models.
•
Informationissecure
The level of protection from data breaches
(encryption, redaction, security, and monitoring) matches the
governance requirements of the data.
•
Informationisdocumented
The information's source system, and all
of the governance rules and transformations that were applied to it,
must be tracked, explainable, and made visible to end users.
Sometimes folks refer to this factor as
end-user transparency
, because
every governance rule and process should be documented and
surfaced to the end user to assist in establishing trust.
Do all six factors of governance need to be applied to Big Data? It de-
pends on the use case: we'll talk about two of them that illustrate entirely
two different governance approaches. For example, analyzing Big Data to
detect fraud patterns may certainly require documentation of the origin of
the data: perhaps it may involve standardization and matching to cleanse
duplicate records, understanding holistic master customer data to match to
fraud records, security to mask sensitive data, and even lifecycle manage-
ment to archive or retire individual records at different intervals. At the
same time, using Big Data to investigate customer sentiment through social
media requires an entirely different treatment of governance. It's likely to
involve integrating master customer data with Big Data to identify custom-
ers, but the social media data may not require cleansing, documentation, or
individual record lifecycle management—the entire dataset might even be
deleted at the conclusion of the analysis.
