Big Data is becoming more meaningful with the ever more powerful data technologies, which enable us to derive insights from the data and help us make decisions. Big Data also creates new courses and professional fields such as the data science and data scientist, which are aimed at analyzing the ever growing volume of data. Some might think this exaggerated because data analysis, after all, not a new invention. However, we might all agree that the progress of digitization associated with the generation of ever larger amounts of data have totally changed the ways we deal with data.
Approaches for data analysis
After collecting the data, you will be faced with challenges to clean and analyze it with appropriate methods in order to derive meaningful conclusions from the data. For example, you will need to structure the individual data processing and analysis steps and to automate and eventually provide the results for implementation. Without the right data analysis methods, tools and other necessary resources, you will not be able to exploit the potential knowledge in the data.
Data analysis in the enterprise can be roughly divided into three types:
- Descriptive analysis – what has happened?
- Predictive analysis – what could happen?
- Prescriptive analysis – what should we do?
These three types of analysis have one common interest – to enable a better understanding of the events, a business, to make the best strategic choices, and as quickly as possible.
Descriptive data analysis
Descriptive data analysis usually materializes in companies as traditional business intelligence reporting, and they usually produces tables and graphs such as revenues by regions or products that describe the current status and their convenient reference and compaction to provide an overview of the current and past situation vividly.
Descriptive analysis is the first analytical phase for business. Most of the analytical tools used by organizations today follow this model. It is the analysis of historical data to understand what happened and why it happened. Thus, companies can identify the reasons for their successes and failures over a defined period, and then reapply the decisions that worked best in the past.
Today, when we talk about data analysis, we primarily denote the descriptive analysis. The use of descriptive analytics is to answer a simple yet crucial question – looking at the data, tell us what has happened, and then, we use our common sense and data mining tools to extract data from different measures or performance indicators (KPIs). These are generally used in dashboards and other software interfaces, and are intended to save time for monitoring or optimization.
When being well conducted, descriptive analysis can bring many benefits to users, whether to aid strategic decision making or understanding the market. But to do this, we must choose right indicators to create great data storytelling.
Unfortunately, in most cases, options for descriptive analysis is often not fully exploited, leaving the user with too many indicators, some less interesting than others, without guidance in their interpretations. Thus, the use of descriptive analytics may create results that are too complicated for data user.
Predictive data analysis
Predictive analysis goes beyond descriptive analysis. It is more demanding than the classic descriptive data analysis in its application – with the help of complex mathematical methods that correlate factors in historical data; for example, characteristics of customers, products or processes, and defined targets identified and displayed in the models. These models have to be validated in tests so they can make predictions for new cases, or new data.
Predictive analysis is the second phase of the business analytics. It provides estimates on the probability of a particular outcome. To develop this type of analysis, the organization must collect a huge amount of data in real-time and add them to existing ones. The analysis predicts what could happen if you take some actions, but it does not help solve the problem of identifying the best option for your business.
Based on the descriptive analysis of historical data and supplementing with some statistics knowledge and algorithms, it is possible to provide prediction solutions to the data users. It is thus possible to compose more powerful tools to refine the information. We can then provide interfaces or utilities that enable decision makers react more intelligently with the inclusion of KPIs in the descriptive analysis.
To make good predictions, you may not need to be a statistician; many tools have been brought from the field of data science. The creation of predictive solutions is facilitated to the delight of users, developers and the data professionals.
Predictive analysis is powerful but users are left on their own in predicting. If the interfaces or tools were misused, the prediction will be wrong.
Prescriptive data analysis
Prescriptive analysis uses two concepts discussed previously in one main purpose – propose ways of optimization for users to support decision making, enabling them to respond faster and more appropriately to a situation in the decision making process. Prescriptive analysis aims to optimize processes, structures and systems through informed action based on predictive analysis – it is essentially telling you what you should do based on what will happen.
Prescriptive analysis is the final phase of analytics for companies because it goes beyond the descriptive and predictive analysis. Prescriptive analysis not only anticipates the possible outcomes (what will happen), but also the reasons for their occurrence (why it will happen) and the time of their occurrence (when it will happen).
Prescriptive analysis tells decision options for the possible outcomes. For example, whether the organization should take advantage of a new opportunity or rather manage a potential risk. It shows the possible outcomes for each decision option as well.
Companies generate new data every minute. These data can be added to the prescriptive analytical tool at regular intervals, resulting in a revised forecast based on micro and macro events happening within the organization or externally. Thus, the organization gets more accurate forecasts and better decision options.
It must be taken into consideration that although prescriptive analysis is economically more interesting than the other two, there remains much more complex to implement. Its implementation should not be taken lightly because it requires much more time invested in its development than a descriptive analysis. In some cases, a user might not need prescriptive analysis; the user may just need two to three indicators that will facilitate the work.
Although the prescriptive analysis offers tremendous possibilities, it is not flawless. These are the same problems that disrupt descriptive and predictive analytics that may also affect the prescriptive analysis. Many parameters, such as consumer behaviour and purchasing context, are constantly changing in real-time; prescriptive analysis requires the information be consistent at all times. In addition, data privacy is a major concern for consumers.
With the powerful and high capacity computing devices, large-scale data can now be connected to real-time applications to support instant decision making from data analysis results. These technologies create a path to prescriptive analysis. As companies reach greater maturity in terms of data and analysis, they want to do more with the information they have and prescriptive analysis can help them unlock their hidden potentials.