Predictive Analytics – The How’s, Why’s, and Common Pitfalls

Editor’s Note: I have long felt that market researchers need to focus more on making predictions, and less on “pure” insights. If we truly want a seat at the table of management, we need to be able to tell a business where it is heading, not just where its been. This is as much as a mindset issue as it is a skills issue. Many researchers, though, have relatively little knowledge of predictive analytics. In the videos attached to his post, Kevin Mabe gives a nice introduction to some issues around prediction. It’s a good place to start. We’d love to see more posts on the topic of prediction – the methods that researchers need to learn, and case studies on the business benefits achieved

Predictive modeling can be one of the most powerful tools that a research, marketing, product development, or finance team has at their disposal. Unfortunately, like many tools of the research trade, predictive modeling is a mixture of both science and art. If used incorrectly or without a good foundation and underlying structure, a predictive model can lead you down the wrong path, wasting valuable time and money. In these videos, we’ll try to showcase the best way to think about predictive modeling and frame things effectively in the first place. We’ll also show how to ensure that your models are sufficiently robust, thus driving meaningful action.

Why is it important?

This one is pretty easy. If you can build a model that incorporates new data in a statistically significant and robust manner, you can effectively look into the future. You can determine which path is going to result in the highest sales, the best margin, or the most meaningful marketing response. You can take action with statistical confidence.  The introduction of machine learning and artificial intelligence has raised the bar of what’s possible and increased the appetite for predictive modeling in general.

Who benefits and why?

We built these videos to target a market researcher/data scientist working in conjunction with a marketing team, but really, any department and individual can benefit from this type of research. We designed the videos this way for a couple of reasons:

1. They are the type of client that we work most closely with
2. It’s easy to see the return on investment from a marketing perspective.

If you can establish which products will sell the best to certain customers, and which products can drive the most meaningful margin, you’ve got great groundwork started. Combine that with a marketing message designed around that segmented market research, and you’ve laid down the roadmap for action and success.

How do I start?

While daunting, the best way to start is to dive right in.  What is it exactly that you’d like to predict and why? What data may help you predict the outcome in which you’re interested?  What new data might you need? If you had the perfect prediction, how exactly would you implement it? Gather as much information as you can about the topic in question, and then start following the videos. If at any point you start to find that you’re not able to tackle on your own, there are many resources both free and paid that you can use.

Good luck and enjoy!

Videos for Support 

• Hard Math with Kevin Ep. 8 – Predictive Analytics (Part 1)

• Hard Math with Kevin Ep. 9 – Predictive Modeling (Part 2)

• Hard Math with Kevin – Predictive Analytics (Part 3)