Business Intelligence vs. Predictive Analytics
– learn the difference between Business Intelligence and Big Data
– understand the concepts of descriptive and predictive analytics
What’s the difference between Business Intelligence and Big Data? Will Big Data replace BI?
These questions are not trivial to answer, and in this lesson we will focus on its implications for decision making (and forget about the technical point of view!).
We can say that Business Intelligence helps find answers to questions we already have, while Big Data helps you find out questions you need to ask.
So BI is more focused on providing information and answers about things in which we are already experts, and answering a question that was already planned.
In fact, when using a BI tool (such as a report with KPIs), it is something that was previously planned. This means, that with our previous knowledge, we planned some metrics and KPIs based on our expertise that will help us make decisions in the future. So the dashboard will give us answers to the questions we planned, but will fail to explore new approaches or answer new questions.
However, in the case of Big Data, we don’t have easy, well-defined reports and answers as we have in BI. Moreover, we don’t have a single system to implement and manage big data like you do with BI.
With Big Data, we are exploring questions that we had not previously planned on asking.
Thus, BI and Big Data are complementary tools.
Business Intelligence systems still have their place in business, as they provide employees with tools to extract well-designed answers to previously designed questions.
Big Data is nowadays a must in every organization, as most businesses don’t always have neat, well-designed questions. In fact, we don’t always previously know what questions need to be asked.
How can Predictive Analytics improve my organization?
To further understand the difference, think about the following analogy.
When driving a car, we basically look at the windshield, to see what’s coming ahead, and at the rear-view mirror to see what we left behind. Business Intelligence would be the rear mirror, as with BI we are looking at the past events, to make future decisions based on past experience.
The windshield would represent Big Data and Predictive Analytics. Through the windshield we can see what will come next, and this is what predictive analytics is about – predicting future events. Thus, in the latter case, we make future decisions based on predictions of what will happen in the future and not just based on what happened last year or last month.
Let’s use a business example.
A health insurance company is analyzing the accident rate of all their policies. They will use the information from this analysis to understand which customers are healthy and can get a discount in their next year’s fee, and which ones are expensive for the company and will not get such a discount.
We can solve this problem through both approaches: just using BI, and complementing it with Big Data and predictive analytics.
Business Intelligence Approach
A simple BI approach could be to analyze the accident rate of every customer. Next, in the simplest case, we could assume that his accident rate next year will be similar to his current accident rate. If we want to be a little more sophisticated, we can add an extra parameter that takes into account that next year he will be one year older.
Any inconvenience using this approach?
Two customers with an accident rate of $10,000 can be extremely different. For example, Customer A had such a high accident rate because he has a chronic disease. In contrast, Customer B had a $10,000 accident rate because he had a bicycle accident and now he is completely recovered. Though we have assigned both of them a $10,000 accident rate for next year, the most probable scenario will be that Customer A will cost the company that $10,000 while Customer B won’t, as his accident rate was due to an accident.
Big Data and Advanced Analytics
In this approach, we will not only be looking at the past, but also predicting the future. Hence, in this scenario we will not assign last year’s accident rate to the customer, but create predictive models that anticipate the most probable accident rate for each customer.
We would realize that the accident rate of Customer A was due to a chronic disease, and consequently will predict the same or higher accident rate for next year. On the contrary, the model will capture the fact that Customer B’s accident rate was explained by an isolated accident that will not likely be repeated next year. Hence, we would assign a low accident rate for next year to Customer B.
The main difference is that with the first approach, we were limited to look at the past, and based on the aggregate quantities, assign future values assuming things will work in the same way next year.
In contrast, in the second approach, we learnt from past data, captured the underlying patterns, and used these patterns to anticipate future events.