There are four important techniques in business analytics that correspond to the different stages of maturity in the analytics lifecycle. Let’s take a look at them below:
1. Most organizations start their analytics journey by asking ‘what has happened’. The business analytics technique that answers this question is called descriptive analytics as it provides a summary of historical data.
2. Next comes the question ‘why did it happen’. The business analytics technique that helps to understand this is called diagnostic analytics.
3. Building further on the insights gained from diagnostic analytics, business leaders want to know ‘what is most likely to happen in the future’. Predictive analytics techniques help to answer that.
4. Business analytics techniques that help to find answers to the question ‘what should we do in the future’ are called prescriptive analytics.
With this overview in mind, let’s turn to an example: take, for instance, a brand seeking to maximize engagement on social media. The brand manager could start with a report of the social media posts of the past months and the engagement that they have generated. This is descriptive analytics. Next, diagnostic analytics techniques could be used to understand why certain posts generated more or less engagement. With these insights, predictive analytics techniques can be applied in order to determine the likelihood of which post will generate more engagement in the future. Finally, prescriptive analytics will enable the brand manager to decide the best posts for the future, based on all the analysis and insights.
Now, let’s look at each of these four business analytics techniques in more detail.
Descriptive analytics
Descriptive analytics is a preliminary stage of data processing that summarizes historical data and prepares the data for further analysis. This usually starts with data aggregation, which is the process of gathering data from multiple sources with the intention of creating a summary. It is extremely important to gather high quality and accurate data and also have sufficient data to gain meaningful results. Today, any business decision requires customer data, and data aggregation must be carried out in order to collate this from multiple sources and systems.
Data mining is used to identify patterns and trends and extract usable data from the largest set of raw data.
Data visualization is also an important part of the descriptive analytics stage. Data is represented graphically using charts and various other visual techniques in order to help stakeholders identify trends and patterns.
Let’s imagine a CEO whose dashboard is set up with visualizations displaying sales by various product categories. She can click on a particular category and see the sales for all the products within it. Then she clicks on the product which is the highest revenue earner in each category. When she looks at the sales for that product over the last few years, the graph theory shows growth every summer and decline every winter. This is as expected, as consumer preferences show that they prefer to consume this product in the summer.
Then she decides to look at the sales for that product in the northern and southern parts of the country. Now the data is somewhat surprising. It still shows higher in the summer and lower in the winter but while this variation is very high in the north, it is much smaller in the south.
She wants to understand the reason for this difference in seasonal variation between the northern and southern parts of the country. Descriptive analytics have indicated what is happening and now she has something specific that has to be investigated with diagnostic analytics.
Diagnostic analytics
Business managers are concerned with tracking performance against KPIs. Their dashboards incorporate data visualizations and show them what has happened against each KPI, but they often lack the context or explanation of why they happened. Diagnostic analytics techniques play an extremely important role here.
The results of descriptive analytics help to identify those areas that require further investigation or identify anomalies. A manager will want to know why the sales of a particular product category increased although there was no change in the advertising campaign, or why shopping cart abandonment has risen suddenly.
Now the data analyst must find a way to drill down further to provide details and explain anomalies. This may require them to look outside the current data set and integrate data from other sources. A study of correlation and causality is also necessary in order to understand the reasons behind the observations. This requires the application of probability theory regression analysis and time-series data analytics.
Considering the volume, variety and velocity of data generated by most systems today, performing diagnostic analytics manually is no longer feasible. Machine learning is being used increasingly for this purpose. Machine learning programs can recognize patterns and identify the underlying drivers of KPIs. A variety of algorithms are used to determine causes and identify the independent variables that business managers can change in order to achieve the desired results.
Correlation analysis helps discover if there is a relationship between two variables or data sets and how strong that relationship is. The Pearson correlation coefficient is the most commonly used; it captures the strength and direction of the linear relationship between two variables. A value of 1 means that as one variable increases, the other also increases. A value of -1 means that as one variable increases, the other decreases; and a value of zero shows that there is no correlation between the two variables.
This method reduces the chances of human bias and mistaking correlation for causation. Experts are then required to provide their experience and judgment and contextualize the results generated by machine learning.
Predictive analytics
While it is important for business managers to know the past performance as well as the factors that have influenced it, what matters most today is to gauge likely outcomes in the future and plan for them. Predictive analytics techniques aim to determine likely future outcomes by building models based on historical data. Different organizations are utilizing these techniques, which are based on statistical modeling and machine learning, to forecast occurrences in the near or distant future.
There are a number of methods, technologies, and tools that help to build predictive analytics, such as data mining, statistical modeling, mathematical processes, and machine learning. Models can be designed to discover the relationship between various behaviors. This can help to predict the propensity of a customer to purchase a particular item or a loan applicant to default on repayment.
One of the popular predictive analytics techniques is decision trees. A schematic diagram represents different courses of action for statistical probability. The branching method shows every possible outcome of a particular decision and how one choice may lead to the next.
Regression techniques are also used in many business situations, particularly in banking, investment, and other financial areas. These techniques help to forecast asset values and understand the relationship between different variables.
These business analytics techniques that help to forecast and predict have powerful applications in many industry segments. Airlines can use these to set ticket prices based on expected travel trends. Retail, hospitality, and healthcare industries can predict future demand and manage operations accordingly.
Prescriptive analytics
Business analytics techniques that recommend the best course of action based on available data are called prescriptive analytics. This field builds further on diagnostic and predictive analytics, but the focus now is on actionable insights and not just monitoring the data.
The aim of prescriptive analytics is to find the best solution among a variety of choices. Decisions can be based on optimizing the results of likely future events and mitigating risks. Prescriptive analytics provides a model to study these.
Available prescriptive analytics solutions are either rule-based or optimization-based. Some products are focused on specific applications, such as supply chain optimization or manufacturing planning. A number of solutions are available for consumer packaged goods and retailers, as these companies usually have sufficient data, and optimizing customer acquisition, retention, and demand fulfillment are business imperatives for them.
An optimization platform typically includes the modeling platform and solvers. The modeling platform defines the problem and creates a mathematical model. The software programs that provide the solutions are called solvers.
The optimization solver software solves problems using one or more algorithms. It takes the problems defined in the modeling environment, along with data. Various techniques are then applied to arrive at answers.
Some organizations are applying prescriptive analytics technologies effectively to improve the planning of sales production and distribution. Some are making decisions about investing in specific projects by considering the environment and compliance concerns of the future. These are the early adopters — and prescriptive analytics is a field that is sure to grow in the near future.
