Skip to main content

Beating Analysis Paralysis With Regression Analysis

mortgage application regression anlysis

By Samantha Beavers

From real estate firms to fast food restaurants, organizations worldwide are leveraging predictive analytics to inform their strategic decisions – and smaller businesses, seeing its promise, also want in. With a variety of techniques and models to choose from, how can businesses get their analytics efforts off the ground?

According to Basiru Usman, teaching assistant professor for Poole College of Management’s department of business management, regression analysis is a good technique to start with.

A statistical method that determines the relationship between variables, regression analysis is one of the most commonly used predictive modeling techniques. With various types, it has a wide variety of uses – which is one of the reasons Usman enjoys teaching it to Poole College’s Master of Management, Marketing Analytics students.

Applications in the real world

Whether they realize it or not, most people have encountered regression analysis before, Usman explains. Anyone who’s applied for a credit card online has run up against it.

“A credit card application asks the applicant a series of questions – like whether they own a house or rent one, what kind of payments they have and so on. Using that information, the credit card company will classify the applicant and determine whether they’re qualified or not. What’s happening behind the screen is a process called logistic regression,” Usman says.

Often used for application approvals that require binary classification, logistic regression models – which have an S-shape – set a threshold around parameters. A value above that threshold will determine one outcome while a value below it will determine another.

“This same process applies to mortgage applications, as well. What I always tell students is that this analytics technique isn’t just about the classroom or the office – it’s about their life. Understanding how it works will, in some ways, give them the upper hand,” Usman says. “Armed with this information, they can know when they’re in a good spot to apply for a mortgage, for example.”

Data-driven car valuations

Regression models are also behind the used car valuations issued by companies like Kelley Blue Book. Using predictor variables such as the car’s age, Usman says, these companies can determine the car’s price – or response variable.

In a simple linear regression, he explains, the car’s value would depreciate by the same amount each year – with the constant relationship between the variables giving the model a straight line. In real life, however, the relationship between these variables isn’t so obvious. Typically the relationship is nonlinear – with cars seeing the sharpest depreciation in value after the first year and a smaller depreciation after that. 

Leveraging linear regression

In addition to teaching students linear regression with real-world examples, Usman is passionate about helping businesses leverage the technique to make smarter, more strategic decisions. Whether it’s understanding how much inventory to carry or minimizing their marketing dollars, he says, linear regression models can help.

With simple regression analysis, Usman explains, businesses can model the demand for their product based on price fluctuations in order to supply the right amount.

“What you need to do first is collect your data. Note how much product you sold when the price was $10 and the number of sales when it was $12.50 and $15. Then, once you’ve collected your data, just fit a simple line,” he says. “Using that historical data, you can determine that your supply should be in a particular range when it’s marked at a certain price.”

And with a logistic regression model, businesses can target the right customers for marketing campaigns.

“Let’s say you have a contact list and you’re trying to determine which of them to include in your campaign. You want to minimize your marketing costs by not targeting people who are unlikely to purchase, but you also want to maximize your profits by reaching as many potential customers as possible,” Usman explains.

“What you can do is make use of your historical data – like the age of the customers, their locations and what they normally buy. Setting a threshold, you can classify which customers are likely to respond and which ones aren’t – and then plan your campaign accordingly,” he continues.

Getting started

For businesses hesitant about getting started with regression analysis, Usman offers an important reminder: it doesn’t have to be complicated.

“Regression analysis can be as simple as you want it to be – and you don’t have to have sophisticated knowledge before you can use it. So if you’re a business wanting to use data to make improvements, there’s good news – this isn’t far from you,” he says.

All they need, Usman explains, is some clean, historical data, a business problem they’re trying to solve and access to Excel. And for businesses getting started with collecting data, he says, less is more.

“Keep your data records as clean and consistent as possible. Make sure that you’re tracking the same variables consistently over time and don’t get distracted by trying to do too much. If some of your inputs have customer arrival time and some don’t, but include extra information like customer reviews, your data won’t be usable,” he explains.

“Start small and make it clean. Put your customer arrival times in one place and track them consistently, and then you can track your customer reviews in another place. Don’t try to mix them together or you’ll have to do extra work,” he continues.

As long as they do that, Usman explains, businesses can have a regression model up and running in no time.

“It’s as simple as four clicks in Excel,” he says. “That’s all it takes to take your business to the next level.”

This post was originally published in Master of Management Marketing Analytics.