Predictive analytics: the crystal ball of Big Data

Predictive analytics is one of the most important Big Data trends. But what is predictive analytics? How does it fit with the other data analytics concepts? Here are the answers.

The Featured Image of Analytics

The best way to describe predictive analytics is to watch the science fiction thriller “Minority Report”. The movie shows a method with which the Police want to hunt (potential) criminals – Predictive Policing, the prediction of crime. And this is how it works: The software calculates the likelihood of a crime occurring in a given area based on historical data. Next, the officers dispatch patrol cars to the identified region to inspect the situation and prevent a potential felony. This is a real-life (and life-saving) application of predictive analytics. It is about making predictions based on historical data.

But what exactly is predictive analytics? The term is often used in the context of business intelligence, business analytics, data mining, or data science. Other keywords such as descriptive or prescriptive analytics only cause extra confusion.

Let’s have a closer look at all those surrounding concepts.

Business Intelligence and Business Analytics are the Collective Terms

Predictive analytics is considered a subset of business intelligence (BI) and business analytics (BA). These terms are often used interchangeably, although there are slight differences in the used methods. Business intelligence is used to answer questions about the current situation. It is concerned with the events of the past and their effects on the present. Business analytics extends BI’s view into the future and relies on a statistical analysis of the company data. It helps to answer questions about the causes, effects, interactions, and consequences of such events. Also, by tweaking some parameters, it uses scenarios to point out alternative courses of action.

The primary difference between predictive analytics and business analytics is that BA is reactive in nature – analysts observe and act on historical data (e.g. revenues, profits, loss). Predictive analytics, on the other hand, is proactive: it assists companies in actionable intelligence and preventive action, even without human intervention.

Another major difference is that predictive analytics mostly uses unstructured public (e.g. social media) or proprietary data (e.g. from research companies). Business analytics, however, makes use of traditional data sources, such as data warehouses and data marts (structured data).

Most likely, in the future, both concepts will blend together and be indistinguishable.

Predictive Analytics is Easy to Understand but Difficult to Implement

The most common definition is:

“Predictive analytics is a method of discovering insights from existing data sets to forecast patterns and predict future outcomes and trends.”

And in simpler words:

“Predictive analytics is the GPS of Business Intelligence”

It takes the business from where it is located now and navigates through a certain path to reach its goals.

The concept in itself is easy. It has been around for decades. But predictive analytics is not the first step you take on when you plan to set up a data-driven business. You need to clarify your strategy, build infrastructure, and get and maintain the right data.

Before committing to long-term investment, you need to understand the ROI (Return on Investment) of it. The cost threshold for an analytics program starts usually around $1 Million. This is how much your company will need to hire a data science team of four to six members (i.e. data scientist, data engineer, data architect, software developer).

To make the most of this investment, you need to structure your data science tasks, as conducting a data science project requires a proper method.

The benefits of accurate predictive analytics models are overwhelming. Take for example Netflix. Its whole business model is based on predictive analytics. The company uses massive amounts of data to predict which shows and movies users would like to watch. They also implemented a well-refined recommendation algorithm to predict what users will likely want to watch. The later was optimized through a competition called Netflix Prize (if you want to try by yourself, here is the data set on kaggle)

Predictive Analytics Is Not the Last Step

Even if predictive analytics is already an enhanced technique of forecasting, there is more. The Gartner analytics maturity model distinguishes four stages. These stages are:

  • Descriptive analytics – What happened? It deals with the past and tries to understand the effects on the present – This is business intelligence.
  • Diagnostic analytics – Why did it happen? It answers questions about the causes, effects, interactions or consequences of events – This is business analytics
  • Predictive analytics – What will happen? It looks to the future and provides predictions about the likelihood of future events based on data mining, machine learning, and other statistical methods.
  • Prescriptive analytics – What should we do to prevent an event from happening? Prescriptive analytics goes one step further than predictive analytics. It also provides guidance on how to influence a particular trend in the desired direction, prevent a predicted event, or respond to a future event.

Potential Use Cases for Predictive Analytics

Predictive analytics is used in many industries with great success. Banks use it for credit scoring, where they estimate the likelihood or risk of a client not paying future installments of a granted loan. Retailers implement dynamic pricing to increase their revenue. In the airline industry, predictive analytics is used to forecast the maintenance of plane parts. Here, the sensors transmit data about the status of a machine or part of it (e.g. temperature). The software analyzes characteristics of usage and condition from several sources and identifies potential errors. The service team can quickly respond and proactively prevent machine failure by installing a new part or starting sooner the maintenance.

There Are Limits to It

No one can look into the future. No one can know and analyze data from the future. Predictive analytics is a method that helps to predict trends. But it is just that, a prediction. Based on a simplification of the existing, complex world. Therefore, use it with caution.

Moreover, the fundamental assumption in predictive analytics is that behaviors in the past will not change in the future. The models in use are mostly static, stationary. Therefore, the once-derived model has to be repeatedly questioned. Managers and experts should keep asking what the fundamental assumptions were and when they no longer apply.

Despite the huge amount of data (Big Data), the biggest challenge is to find suitable data. To make predictions about the future spreading of diseases, data about the past spread is necessary. Data across all contact points must be consolidated accordingly. Here, you need to set up solid professional information management. It is the prerequisite for successful predictive analytics to turn big data into smart data. Predictive analytics starts with a state-of-the-art information management solution.

I don’t have any particular specialist skills. I have a sort of vague knowledge of many areas.