Predictive analytics is a powerful method used by organizations to predict future events and behavior in order to optimize current operational (marketing, sales, product) decisions. The prediction is based on a fundamental concept that history tends to repeat itself. In other words, past behavior predicts future behavior.
As an example, let’s look at an online pharmacy store. Historical data indicates an 8% contact rate; i.e, of the 1000 website visitors, 80 visitors contact customer service within 24 hours. The vice president of online operations wants to reduce the contact rate because each contact costs operations increasing amounts of money. Let’s hypothesize that a 5% decrease in contact rate would save one million dollars annually in reduced operations cost.. The operations team decides to use predictive analytics to understand the drivers of contact.
By applying predictive analytics techniques on historical data, a relationship is identified between help page visits and visitors making a phone call to the customer service department. In particular, 67% of visitors call customer service after having gone through two distinct help pages. This is a very helpful clue. If the visitor can be intercepted before they hit the second help page, either by providing better help content, a live chat or a clarification window based on what they are browsing, contact rate can likely be reduced. I have seen organizations save and make millions by understanding such kinds of pivotal relationships between behaviors and events.
On the flip side, in spite of being a powerful optimization technique, predictive analytics is often left to the devices of data miners and data scientists and hence is often misunderstood and misused by businesses. So in this article, I am going to talk about some of the common misconceptions about predictive analytics with the intention that once clarified, it can be used more appropriately to drive business decisions.
- Predictive analytics is not the “new kid on the block.” Some of the recent national media reports of predictive analytics application make it sound like predictive analytics is a newfound technique. But, in fact, predictive analytics is not new. In recent history, Fischer and Durand built one of the first credit scoring models back in the1930s. Additionally, I would say that predictive modeling techniques goes back thousands of years – the use of Indian astrological charts in arranged marriages being one such example.
- Predictive analytics does not make perfect predictions. Often while building the model, it is clear to all that model prediction has a probability associated with it, but upon completion, there is often a misplaced sense of perfectness in the predictions. The model predicts by maximizing likelihoods and there is always a certain degree of misclassification. By using other predictors, these odds can be improved, but it will still not be 100% perfect.
- Good predictive analytics software tools do necessarily equate to good models. With tremendous development in the software tools front with better graphical user interfaces (GUIs) as well as higher automation, new users often mistakenly believe that a good model can be built automatically by pressing the GUI “build model” button. But that is far from the truth. Building good models requires proper technical skills and use of a model building process. Though surprisingly, sometimes even that does not deliver a good enough model.
- A good model does not guarantee business results or profit. This is one of those highly prevalent myths that even experienced analysts fall for, often finding themselves frustrated that nobody in the business seems to care for the amazing model they have built. Good models generate business impact, but only when the right stakeholders are brought into the analytics process at the right time, building proper alignment toward actionability using a framework like BADIR - Aryng’s 5 Step framework. In our online pharmacy store example, if instead of building a contact rate prediction model, we had built a model to predict visitors most likely to do live-chat, would the vice president of operations care for the model and the results? Probably not. Unless we can show the relationship between what we (predictive model/team) are trying to do and his business goal (contact rate reduction), the operations team is unlikely to use a perfectly “good” model.
- Like relationships, models once built need to be nurtured. Models get stale over time and if not maintained, often stop delivering the incremental value it started with. As organizations embark on the journey of competing on analytics, they need to be aware that it is not a one-time investment. You can’t hire external consultants, get the model built and leave it at that. Models need to be tested, tweaked and maintained to continue delivering the incremental benefits.
To learn more, you can download the whitepaper “The Five Myths of Predictive Analytics”
and/or attend one of our upcoming workshops, Business Impact Through Analytics
, May 18 in Santa Clara, CA.
SOURCE: The Top Myths of Predictive Analytics
Recent articles by Piyanka Jain, CEO of Aryng