Data analytics, as defined in “Competing on Analytics: The New Science of Winning” by Thomas H. Davenport and Jeanne G. Harris, refers to the extensive use of data, statistical and quantitative analysis, explanatory and predictive models, and fact-based management to derive decisions and actions. It has the meaning of discovery and communication of meaningful patterns in data.
Dealing with business analytics implies the efficient use of quantitative analysis, statistics, as well as information modelling to shape business decisions. In this context, people dealing with business analytics can be classified into three levels: analytics scientists who build complex models to extract insights from data, analytics experts who apply the models from the first level to real business problems, and analytics specialists who can build insights based on the output of the previous steps.
Typically, organizations start their journey into the realm of business analytics by examining the data generated from the different systems. For organizations, data can be: First, Structured, which means it can be stored in databases (in tables) and can be accessed through database management systems. Second: Unstructured, which cannot be stored in traditional relational databases. Unstructured data come in formats like text, image, video, document, etc. It can also be in the form of customer complaints, contracts, or internal emails. The importance of unstructured data lies in the embedded interrelationships that can never otherwise be discovered. Both structured and unstructured data establish the full scope of information needed for improved decision making, and many of the analytics techniques that are used here can equally be applied there.
While the term “business analysis” is sometimes used interchangeably with business analytics, the two terms have distinct implications and there are some differences between them: Business analysis is more related to functions and processes, and is meant to recognize business needs and determine solutions to particular business problems. Business analytics, on the other hand, implies the use of tools, techniques and skills to investigate the past business performance and build insights for future. Yet, both of them aim at improving businesses by measuring the business requirements and finding solutions to problems.
Unfortunately, not all organizations know exactly where to start on analytics, and what types of analytics would benefit the organization more. Four types of analytics need to be considered for a better leveraging of the information that the organization has. The value behind them is that they can ensure a better engagement with customers. The four levels are:
- Descriptive analytics. This is the most common and most well-understood type of analytics because it covers the most business needs. It is commonly referred to as the simplest type. Historical data is used to understand and analyze past business performance. The past here can mean from one minute ago to few years back. It can also be used to understand events in real-time. The significance of this type is that it helps understanding the relationship between products and customers which will help take actions in the future. Descriptive analytics include process such as dashboards, reports, and queries. Most processes can be implemented through on-the-shelf business intelligence applications or spreadsheet tools. One example of descriptive analytics is examining historical electricity data in order to put power plans and set prices.
- Diagnostic analytics. This type looks at past performance and tries to understand what happened and why happened. Trying to dive into the data, diagnostic analysis is used by businesses as a root-cause analysis to identify the factors that affected the bottom-line. One common example for diagnostic analytics use is in social media applications where the service administrators want to assess the factors that assisted in past campaigns and what did not. However, the work included in this type is very laborious and only a small portion of companies implements it with the aim of gaining a good understanding of a small piece of the problem.
- Predictive analytics. Using statistical models, machine learning, data mining, genetic algorithms and game theory, this type of analytics can help making predictions about future events by finding trends and behavioural patterns located in data. For example, it can be used to predict risks and opportunities in the future. When in process, predictive analytics does not give a single future-state but rather a number of future-states with the probability of each of them. Common example use of predictive analytics include the study of the likelihood of customers making future payments on time, or forecasting the demand to adjust production for a certain region or customer segment. However, because the tasks incorporated are not easy to implement, only a small number of companies use this type of analytics.
- Prescriptive analytics. It is a complementary step to predictive analytics because it plots the best courses of action in response to what predictive analytics tells about the future. It also includes the identification of likely outcomes of each decision or course of action. For example, drug marketing companies use perspective analytics to improve drug development, reduce marketing time for new drugs and find the right patients for experiments. In fact, the real advantage of analytics comes with this stage, as it tries to answer the question “Now what?” or “So what?”… To do its job, it applies a combination of different tools such mathematics, business rule algorithms, machine learning, and modelling over a range of historical, transactional, public and social datasets. Two primary approaches are applied here: simulation and optimization. However, as pointed out by Gartner, this type is still in its infancy and it needs more time to be fully mature. Hence, only a small percentage of organizations use this type of analytics.
It is not always possible to say that the “descriptive analytics” is the easiest step to implement as it all depends on the data we have, the level of analysis that we want to implement and the analysis tool that we are using. Also, the order of processes does not always have to start with descriptive analytics. For example, if we know what the problem in the data that we have is, we can skip this step and go straight to “diagnostic analytics” or “predictive analytics”.
Finally, none of the four business analytics types can be described as better than the others. The type of analytics to be done depends heavily on the issues that need to be addressed, the importance and the ease of the project, and the skills that the team has. An optimal business solution requires the implementation of the four steps all together. Although they are different, they are necessary to provide a full view over the performance of the organization.