Data Warehousing, BI and Data Science

27 June 2018

Can Machine Learning replace BI?

Filed under: Business Intelligence,Data Science — Vincent Rainardi @ 6:50 am

In the last 20 years companies have been slicing and dicing data, looking for patterns in the data manually. They then use the patterns that they find to make business decisions. But in the last 2 years companies have been using machine learning to find the patterns in the data, and use it for predictive analytics.

Business Intelligence

This is an example of a typical BI dashboard for sales, source: https://www.cbsi-corp.com/software/business-intelligence-reporting. The dashboard presents the sales data. On the top it is sales and profit over time and by product. And on the bottom it is sales by salesperson and profit by customer segment and product group.

Example of BI Dashboard

Based on this data the company can make business decisions such as:

  • Increase or decrease the profit margin for a particular product group.
  • Focus the marketing effort on a particular customer segment in a bid to bump the sales.
  • Restructure the sales teams in order to improve sales performance.

So BI is supporting the management to manage the business better by making them better understand the current and past business situations.

Machine Learning

A typical machine learning is fraud detection. Thousand of transactions a day (say credit card or payment/bank transfer) are analysed by ML algorithm and a few which are suspected are automatically put on-hold and manually checked. Date, time, amount, customer profile, payee, and transaction/payment type are “features” of the neural network algorithm which scores each transaction, and transactions which over certain limits are put on-hold.

In this case ML is replacing BI. The ML system automatically does the actions, which is holding off the transaction, and putting it in the queue for manual review. The fraud analyst then review and check with the customer. They then release the transaction if it is genuinely initiated by the customer or block the transaction if it isn’t. BI can’t do this. We need ML to do this.

But can ML do the sales dashboard above? No it can’t. ML is build to analyse the current situation, but it can’t suggest that Corporate Gift is the one which we should focus our marketing effort on, not Mugs and Drinkware. There are a lot of information which are not included in the data, and it is this information that enables the business users to be superior than any ML algorithms.

Can ML replace BI?

So can ML replace BI? Only in certain areas I believe, but not in its entirety. As I wrote in my book there are 3 areas of BI: Reporting, Analytics and Data Mining. It is the Data Mining area which is being replaced by ML. But the reporting and analytics are not.

Advertisements

Create a free website or blog at WordPress.com.