Definitions, and how they apply in analytics
With the rise of technology applied to decision-making, many terms have appeared that are now in fashion — among them Business Intelligence, business analytics, big data, analytics, artificial intelligence, machine learning and deep learning.
Understanding how these terms apply in practice can become a tangle, because there is no specific consensus on their definition. Fortunately the management user has some sense of what they mean at a macro level; in the lines that follow we will fill those concepts out and set out what they are worth to the company.
Analytics is the exercise of mathematics with criteria of analysis — of discernment — for the purpose of making decisions. Under that definition it is clearly not a new concept: it has been applied in civilisation for more than five thousand years. The result of applying analytics is a discovery — an insight — that makes intelligent decisions possible.
Technology has improved analytical processes drastically over the last 15 years, through two capabilities:
Processing speed
Data storage
We are now in a new industrial revolution: the age of data.
The terms we began this article with slot into the types of analytics:
1. Descriptive analytics (Business Intelligence, business analytics): focused on controlling business variables through dashboards. It is framed in the past and answers the questions what happened and why it happened. It aspires to have information live. It has four kinds of user:
Management: reads the information on dashboards developed through management concepts in order to automate decision-making. It starts by setting alarms and extends its reach to take in every business question the manager has to answer in order to manage. It lets managers focus on planning, creativity and optimising efficiency rather than on directing and policing information.
Analyst: consumes information through technology connected directly to the company's data sources, modelled into information cubes.
Systems: maintains the hardware and software that supports the solution. Assigns users and deploys cubes, and generally does not take part in exploitation and analysis.
Predictive analytics: looks at the future within a static set of data, applying probability and statistical models. It generates future scenarios for management to analyse.
2. Prescriptive analytics: looks at the future within a dynamic set of data — a constant flow. It builds models and algorithms that learn through comparison (machine learning; deep learning is a kind of machine learning that uses neural networks as its base learning model).
3. Autonomous analytics: the same as prescriptive analytics, but the models and algorithms learn by themselves without human intervention (artificial intelligence).
From the above we can infer that Business Intelligence and business analytics are the same thing. Big data falls under the same definitions but with a greater quantity of data — a great deal more, above 100 terabytes. What distinguishes big data is that it can also be applied to prescriptive analytics models.
It is advisable to have a descriptive analytics scheme (Business Intelligence) deployed first before applying predictions, since the resulting prediction has to feed back into a decision process in order to be useful to the company.
Fortunately implementation costs have changed too. When this new era began, implementing an analytics project could cost hundreds of thousands of dollars; today the reduction can reach a tenth of those figures.
