3 Questions Predictive Modeling Can Answer for Marketers | - Digital Advertising Agency – Marketing Strategy | IQ Agency 3 Questions Predictive Modeling Can Answer for Marketers |

3 Questions Predictive Modeling Can Answer for Marketers

Predictive modeling has become the forefront of analytics and has changed the way companies look at consumer data. Broadly defined, predictive modeling is the ability to predict future behaviors using models that are empirically derived and statistically valid. For marketers, predictive modeling uses past consumer data to predict how your marketing efforts will affect consumer behavior and decision making.


To better understand the benefits of predictive modeling, we look at three questions predictive modeling can answer for marketers.

Which customers should I focus retention programs on?

Suppose you have two customers, Customer A and Customer B, and you want to know which one you should target with your customer retention program. Customer A was referred to your company’s website from organic search results, spent more than $50 on their first transaction, is male, and is a millennial. Based on past data, your model predicts this customer has an 80% chance of becoming a repeat customer organically. On the other hand, Customer B was referred to the website from a Facebook ad, spent more than $50 on their first transaction, is male, and is a Gen Xer. According to your predictive model, this customer only has a 10% chance of becoming a repeat customer organically.  Since Customer B has the lowest chance of becoming a repeat customer, you should use the customer retention program to try and convert them.

Predictive Modeling | Choice | IQ Agency

How is my social media page affecting revenue?

Your company started a Facebook page 2 years ago and is struggling to realize its impact on revenue. You decide to use a predictive model to better understand the ROI of your Facebook page and to find out how many Facebook interactions you need to realize target revenue. Your independent variables are monthly engagement rate and reach and your dependent variable is monthly revenue.  By comparing data from past months, your predictive model tells you that a 1% increase in monthly engagement results in a 10% increase in revenue, and a 1% increase in reach only results in a 0.05% increase in revenue. This data shows you that even though you may be reaching a high number of users, revenue will only increase if users are engaged.

Now we can also predict how much engagement and reach we need to realize X revenue. Suppose average monthly engagement is 8%, average monthly reach is 100,000 Facebook users and average monthly revenue is $50,000. Your CEO wants to increase average monthly revenue to $55,000 and wants to know what average engagement rate and reach should be to reach that number. With a little math using the numbers you have, you find that an average monthly engagement rate of 8.8% and an average monthly reach of 110,000 Facebook users will lead to monthly revenue of $55,000.



How can I build a personalized experience for the customer?

A few years ago, our VP of Strategy, Noah Echols received a strange package from Target. Inside the package was a free container of baby food and coupons, but Noah didn’t have any children and was only in preliminary talks around the subject with his wife. Even though Noah didn’t have any kids yet, Target knew that he was planning to by using predictive modeling techniques. Target and other big companies can use the vast amount of consumer data at their disposal to predict life-altering moves such as having a baby. In Target’s case, their REDcard gives customers savings in exchange for data that would usually come from a third-party data collector. The theory behind this is that if a company can correctly predict life events of their customers by using past purchase data, they can build a personalized experience for a customer that they couldn’t find anywhere else. (see our post on The ROI of Customer Experience)

The affect of predictive models on business outcomes is wholly determined by what types of data you are comparing and what type of model best captures the features of that data. When building your model assumptions will be made, however, they can be hedged against through proper monitoring and testing to ensure valid insights. With a proper model in place, companies can predict behavior and decision-making patterns and in some cases start trends themselves. Predictive modeling is changing the way marketers look at consumer behavior and the future is bright for those who have already implemented these approaches into their marketing departments.

Leave a Comment

* Denotes required field. Your email address will not be published or shared.