Digital Profiles Part N: Statistical Modeling

Alert: This post is for statistical computer geeks ... people like me!

Digital Profiles are an absolutely perfect complement to the statistical modeling process.

In this simple example, I am using Logistic Regression to predict the likelihood of a customer purchasing in the next month, given how much the customer spent in the past year (divided by 1,000) and the Digital Profile that the customer belonged to.

Fifteen of the sixteen digital profiles are listed ... the sixteenth digital profile (Rotary Phone) is assumed to have a value of zero.

Three Digital Profiles have coefficients greater than one, suggesting that they significantly contribute to future response (Gold Mine!, Multi-Channel Mavens, Retail Fanatics).

Digital Profiles with a negative coefficient are those that deflate future response.

I use Digital Profiles in my Multichannel Forensics projects, as well as my Online Marketing Simulations.

1. Digital Profiles are combined with historical recency and historical spend by year to predict the probability of a customer purchasing.
2. Digital Profiles can be used to predict how much a customer will spend if the customer chooses to purchase again.
3. Digital Profiles are THE MOST IMPORTANT DRIVER in calculating the "ORGANIC PERCENTAGE", the percentage of future demand that will happen catalogs are no longer mailed to customers. Digital Profiles do a very nice job, especially when your organic percentage is close to fifty percent.

It's been my experience that, after recency/demand, Digital Profiles are the most important predictor of future response, future spend, and future organic demand.