Using Predictive Analytics to Predict Customer Behavior

Web Analytics and Predictive Analytics

In Chapter 3 of Web Analytics 2.0, Kaushik (2010) discusses diagnosing the root case of a metric’s performance by using predictive analytics. He gives an example using an ecommerce website that wants to improve its conversion rate by 10%. Kaushik (2010) recommends going through a “root cause diagnosis exercise.”

Before figuring out how to improve the conversion rate, all the variables that influence or have an effect on the conversion rate must be identified. Examples of these variables are acquisition strategy, organic search keyword ranks and the ease of your checkout process. Once all the influencing variables are identified, data must be collected for each of the variables. Than all the variables must be analyzed to determine where the true opportunities are for improving the ecommerce site’s conversion rate (Kaushik, 2010).

According to Kaushik (2010), the output of this exercise will be something like this: here are three areas where the ecommerce site stinks. Then a cost-benefit analysis can be done to figure out which areas will give the maximum bang for your buck.

This is a great example of how using predictive analytics can help market a business. There are numerous other ways that predictive analytics can be applied to marketing situations to help marketers make better decisions and communicate more effectively. The one I will discuss today is customer profiling and segmentation.

Predictive Analytics and Marketing

Wouldn't every business like to know which of its customers are the most profitable? And what their characteristics are so more customers like them can be found? Of course they would, and this is why marketers are so interested in predictive analytics for customer profiling and segmentation.

According to an SAS whitepaper, classification trees or logistic models can be used to help companies understand their customer base by segmenting customers based on measures that matter to the business, such as response, revenue and risk (Parr-Rud, 2012).

For example, a technology company is preparing to launch a campaign for a combination printer/fax/copier targeting a list of current business customers. To group customers into three distinct groups, two models were created. A logistic regression model predicted the likelihood that the customer would buy the printer, and a linear regression model estimated the amount the customer would spend, given that a purchase is made. The product of these two scores gave each customer an expected value. The company then divided its customer list into groupings that allowed marketers to create a strategy based on the customer’s expected value (Parr-Rud, 2012).

In addition to better targeting customers for their new product, the company also used the firmographic profile of the highest value customers to identify new target markets for acquisition marketing (Parr-Rud, 2012).

This is a few of the ways predictive analytics can be used to determine and predict customer behavior.

References

Kaushik, A. (2010). Web Analytics 2.0 The Art of Online Accountability & Science of Customer Centricity. Wiley Publishing, Inc.

Parr-Rud, O. (2012). Drive Your Business with Predictive Analytics. SAS Whitepaper. Retrieved from http://www.sas.com/reg/wp/corp/42596