Most raw data, and particularly Big Data, are not fit for human interpretation and do not provide significant value at their unprocessed form. Descriptive Analytics aim to produce understandable chunks from the initial data, allowing people to better interpret them and deduce useful information for their uses and purposes. They are used by 90% of organisations and business today and are the foundation of Business Intelligence processes.
Descriptive analytics are applied on historical data of variable timeframes, and essentially summarise what has occurred in that timespan. The perspective that are to be evaluated are determined by the decision maker and are usually translated as calculations over different data dimensions.
Types of Descriptive Analytics
Common actions like classification and clustering, filtering, aggregation and statistical analysis fall under this category of analytics. Descriptive Analytics are therefore used when there is a need to understand data at a coarse grain and when there is a need to summarise and describe them from different aspects.
Of course, such analysis of the past, alongside its inherent value for assessing a current state, is the driver for understanding how past activities and results might influence future outcomes.
The automation of the latter is achieved via a different class of analytical process, Predictive Analytics
As Descriptive Analytics, Predictive analytics operate over historical and recent data. However, their target is not to summarise the existing data to more actionable information. Rather, they attempt to predict data that are not actually available, either because they are missing or because they refer to the future.
To do this, Predictive Analytics use various statistical and machine learning methods to quantify the likelihood of a future outcome based on current trends and patterns.
Types of Predictive Analytics
- Predictive Modelling: What will happen if certain conditions hold?
- Root Cause Analysis: Why this actually happened?
- Data Mining: Identify correlations between data
- Forecasting: What will happen if the existing trends continue?
- Stochastic Simulation: What could happen and how probable is each possibility?
- Pattern Identification & Alerts: When should actions be undertaken to correct a process or sustain an outcome?
The capability of predictive analytics to clearly showcase probable outcomes and identify critical turning points, is the standing ground for the next evolutionary step of analytical processes, where computational methods are actually able to propose solutions for critical situations and optimise processes and approaches for the problems at hand. This concept of advanced analytics is known as Prescriptive Analytics and it is the point where data and advanced computational models lead to truly automatic, highly intelligent decision support.
Prescriptive Analytics share with Predictive Analytics their stochastic interpretation of the future, based on the analysis of already available data. Their crucial difference, which leads to their nomination as the “final frontier of business analytics” is the fact that they aim to determine “why” the predicted future outcomes will happen and offer ways to manipulate these outcomes. Effectively, it answers to the question “What should we do?”, by having analysed and rationalised “where do we stand?”.
Prescriptive analytics use various mathematical and statistical models coming from different fields like natural language processing, machine learning, operational research, etc. These are combined with different business rules related to goal setting, best practices, legal obligations and constraints, etc.
Ultimately, Prescriptive Analytics allow the mass-scale simulation of scenarios, are able to determine the effect of different decisions for a given process and ultimately, provide suggestions for the optimisation of the process with respect to the defined targets. They are, therefore, the most explicit, effective support mechanism for the decision maker.
