Too “Logit” To Quit: 5 Ideas For Using Logistic Regression In Your Business

Predicting behavior is the “white whale” of every business leader responsible for business development and customer acquisition. The field of data analysis offers numerous techniques and methods to help business leaders in these efforts, yet for those without the experience and/or resources this can be a daunting task. In this article I want to discuss one technique that can support the efforts of any business, small and large: logistic regression.

What is a logistic regression?

A logistic regression is a type of regression model that quantifies the odds of an event occurring. It is particularly useful for measuring the odds of binary data (yes or no). This statistical method has applications across many fields from medical to social sciences and can uncover great insights when applied to businesses.

Logit = L = Constant + a*v1 + b*v2 + … x*vn

The logit is the binary variable being regressed (buy/no buy, yes/no, etc.) represented as 1 or 0. Each explanatory variable (v) can be binary or continuous value that predicts the logit. The coefficient for each explanatory variable in the model represents the amount of change in the odds of the logit event occurring for a change of 1 unit.

The simplicity of the logistic regression versus other methods makes it a popular method for analyzing binary data. It is simple to interpret yet has powerful capabilities that any business can apply with the right data. Below I will discuss 5 ideas of how you can use this method to obtain insights about your business.

1. Customer Quality Scoring

Your business has a wealth of information hiding in historical prospect and customer data. In direct marketing efforts through mail, email, telephone, etc. you have collected valuable data that can help you understand which customers are likely to buy. By regressing on prospect characteristics such as demographics, geography, etc. you can quantify the probability that a customer will buy based on these explanatory variables. This is in effect creating a quality customer for each prospect to identify those with highest potential for successful conversion.

2. Campaign Analysis

Analyzing the effectiveness of multiple campaigns can be a complicated process. The logistic regression provides one method for understanding the performance of each campaign. With a list of customers, customer-specific characteristics such as a quality score, and inclusion in one or multiple campaigns, you can calculate the odds that a campaign drove the sale.

3. Sales Attribution

This is essentially the same analysis as discussed above but with a different question. Rather than trying to determine if certain campaign(s) are effective at converting customers, in this application we want to calculate the contribution of each marketing channel is to a sale. Sales attribution is a major obstacle for businesses trying to understand the true incremental impact of marketing channels. The logistic regression offers one way to overcome the flawed first-touch or last-touch models that are predominantly used today.

4. Customer Retention

As the saying goes, it is cheaper to keep an existing customer than it is to find a new one. Wouldn’t it be helpful to understand which customers are more likely to stay and which ones are more likely to leave? The logistic regression can help you gain these insights about your customer. With this information you can find out the drivers that lead to customer retention and can be used in your back-end marketing, or customer nurturing, efforts.

5. Credit Risk

If your business extends credit to customers then you understand the importance of collecting on outstanding debts. A logistic regression can be applied to identify whether a customer has an elevated risk for default. For example, setting your logistic category variable to “default” or “no default” and modeling on explanatory customer-specific (age of person/company, size of company, prior purchases, etc.) and macro variables (unemployment rate, GDP, etc.) will quantify the likelihood that a particular will default. This will allow you to make better credit decisions, such as whether to extend credit or not and the terms.

The number of applications for the logistic regression can go on and on from here. But hopefully this short list serves as a starting point for understanding how to use this tool in your decision-making process. If you see the value, then the next logical question is how can you implement this technique with the resources you have available. Fortunately, the R statistics package is a free tool that can be added to your data analysis capabilities. Learning to use this tool or with the help of a data analysis professional you can discover important insights about your business.