Do Predictive Analytics Yield More Value Than BI Systems?

Line Art

The short answer is yes. Predictive Analytics add layers of decision analyses that, while supported by BI (Business Intelligence) data-mining technology, deliver far more strategic value than BI by itself.

The Process

Before getting deeper, let me say that both Predictive-Analytic and BI technology platforms are valuable, even essential, tools in today’s competitive landscape. That does not mean they are to be valued equally. Let’s start by defining each process.

  • Business Intelligence: Traditional BI systems are data-centric. They access centrally stored transaction records and employ data-mining and data-exploration techniques to create reports, dashboard user interfaces, and alerts. Some even identify trends and patterns, and apply statistical algorithms to extrapolate the past. In general, however, BI exercises are primarily focused on mining history and presenting and analyzing what has already happened and why. BI commonly focuses on questions of how much, how many, how we are tracking to goal, and which records stem from anomaly events. All very important, but not a clear picture of the future.
  • Predictive Analytics: Predictive analytics is quite different in that it is knowledge-centric rather than purely data-centric. In addition to finding and extrapolating trends and patterns in historical data, Predictive Analytics capture management estimates and insights from throughout the organization and merge them with historical data for high-level modeling and simulation. Predictive Analytics predict future events, actions, and behaviors, such as a person’s buying habits, or the behaviors of an individual, group, or commercial entity (such as a store). Through advanced techniques such as Monte Carlo simulation, Predictive Analytics can tell us not just what might happen in each competing course of action, but how likely it is to happen. These exercises are intently future-focused and applied to making optimal business decisions.

Let’s further differentiate Predictive Analytics from BI using the example of a store. BI can provide demographic and behavioral details about each customer, such as home address, activity history, age, communication preference, income bracket, and more. At a more aggregate level, BI can also tell us the store’s average shopper demographics, or which stores in the chain cater to which subsets of shopper demographics. This information provides substantial value in telling us which products sold, where, and to whom; or which promotions worked, which did not meet expectations, and so on.

By contrast, when we view this valuable history through the lens of Predictive Analytics, we can also identify customers that will buy next month and which products will sell. From this, we can determine which products to stock, which promotions to fund, which locations to expand, and which customer experience to invest in. We can also compare and contrast the value and likelihood of outcomes for these and other competing courses of action.

Additionally, we can analyze outcomes based on myriad what-if scenarios. This can tell us which promotions will fare better next month given the time of year, the possibility of extreme weather, or other demand-driving events. Or, we can determine which products to stock up, which to keep level, and which to pause or drop based on known competitor actions, new product introductions, or myriad internal and marketplace scenarios. From there, we can compare and contrast each course of action by what it portends for revenue, liquidity, staffing requirements, and much more.

In sum, we can see that both Predictive Analytics and BI are important to understanding and improving business operations and that both have their place in the analytics framework. However, only one (Predictive Analytics) is a truly future-focused strategic decision tool. With that, I believe that Predictive Analytics delivers more true value to the organization and marketplace than standard BI. I also believe that BI has amassed considerable perceived value, perhaps because it is often conflated with Predictive Analytics.

Knowing what happened and why is a valuable first step in helping us understand the past in terms of what sold, where, and to whom. However, this is not nearly as valuable as knowing what is going to sell, with what certainty, through which course of action, and under which scenario. That is the wisdom that will drive maximum value next month, quarter, and year.

I will leave you with a question: If in one hand I had the winning numbers to last week’s lottery drawing and in the other hand, next week’s winning numbers … which would you choose?