Analytics technology has already become a commodity. Analytics technology remains a necessary condition for a data-driven enterprise but having Analytics technology is not a sufficient condition for a data-drive enterprise.
The rise of Hadoop signifies that Enterprises are paying equal attention to both structured and unstructured data. There are many good developments around Analytics technology but it seems it would take a while for the end user community to start thinking beyond just data visualizations. And, that brings us to a fundamental question with which I would like to lead this article.
The Analytics sales and marketing talk can play a huge role in shifting the semantics of Analytics beyond the popularly accepted Analytics outcomes i.e. Dashboards & reports. And therefore, I think we can ask ourselves the below question.
Does your data have shape?Interesting question, isn't it? But to really answer this question we have to first look what analytics is really. Is Analytics limited to technology tools like ETL or Business intelligence or the Hadoop. Does technology completely explain Analytics?
The answer to the above question is "No". In fact, Analytics has a lot to do with science than technology. Both technology and science enable Analytics. What scientists did in their research papers in the pre modern computer age was also Analytics.
Coming back to the our question, Whether data has shape? We know for sure that physical objects have shape. But is data physical? Can we claim that data is material entity?
The point i am trying to approach is that I see a lot of users have developed a certain mindset that Analytics means Dashboards and reports. I think a part of the reason for this to happen is a deep innate human craving for finding shape in things.
No wonder, eyes are the strongest sensory organ in the human anatomy. But we should also be willing to consider the fact that this tendency sometimes does limit our ability & capability to interpret data and truly become data driven.
Data is a abstract thing. Data is a encoding mechanism, it encodes reality. Its certainly not physical as its merely a representation. While data does not have physical shape we can argue that it has geometric shape. But, do we really need to focus so much on the geometric shape?
In some cases, yes, for e.g., when we look at the dial gauge on your car, that's a type of analytics called operational reporting. Clearly, there are some types where the geometric representation of the data is important but that should not be only method.
Arithmetic techniques are way superior to Euclidean geometry in data analysisThe most obvious limitation of Euclidean geometry is that it cannot handle nominal/categorical data. Arithmetic techniques suggested by Newton, Fisher, Bayes and others are far more superior in analyzing data. This will remain undisputed for many years to come.
The rise of a new segment called Topological Data analysis is analogous to the rise of Visual discovery and search based Business intelligence tools. A limitation of a class of mathematical techniques that depend on Euclidean geometry like Topological Data Analysis is that they cannot be used for scoring.
They can help in the explaining the dataset to some extent but not predicting data. Any such models cannot be operationalized, for instance, like a recommendation engine i.e. put in a production system in lights off mode.
Machines don’t need data visualizations
Internet of Things (IoT) is expanding into more and more industries and is one of the most significant contemporary movements.
Let’s accept the fact that more and more aspects of decision making are getting automated and automated analytics depends on visualization only for performance monitoring and not for the actual act of taking the decision.
Let's begin with the end in mind
All investments in Analytics are truly to improve the decision making abilities of decision makers within an Enterprise. Data is collected via instrumentation techniques for the same purpose. Pre built dashboards and reports are fine for the ERP data or for operational reporting. Stopping with Operational reporting is also means not completely benefiting from the insights held by the data.
A lot of companies have data, they might not have certain important data but there should be a good review of the data requirements of an organization. Sometimes, one set of data carries the effects or the shadows of the missing data and such equivalent data can then be used in the Analytics exercise. The mandate for Analytics process owners should be
1) decisive insights,
2) top class data quality and
3) decision making governance.
A lot of data generated within organizations today is structured data, some organizations are also generated huge amounts of unstructured data for e.g., QSR chains have ordering portals and mobile applications. All these apps and portals are generating tremendous amounts of data. A lot can be done to with this data beyond the realm of a Dashboard and report. I would like to advocate the heavy use of Arithmetic techniques on the generated data.
Guest Author
Sangram Aglave brings a unique perspective on topics related to Enterprise IT Applications given his diverse professional experience in all functions of the Enterprise IT Applications business like Sales, Product & content marketing , Project management & Software product management. He is a ex-Oracle Business Analytics product manager and has worked at various silicon valley based product startups.
