Math Meets Marketing with Statistical Models

Statistical modeling, a simplified, mathematically formalized way to approximate reality and then make predictions based on that approximation, has been around in some form since the 1970s, but it is only now starting to find an important place in digital marketing as a way to help companies predict human behavior.

Thanks to advances in technology and the accessibility of huge amounts of data, statistical modeling is more viable for businesses today, but there is still a lot of mystery around what it is, how it works, and how the models are created.

“Statistical modeling is a way to understand the real world using math,” says Nick Ziech-Lopez, director of product strategy at MessageGears, a provider of cross-channel marketing messaging technologies. “Can I predict what you want from me? The algorithms dictate what I should and shouldn’t send you.”

“Statistical marketing is important for most organizations in general, but critical for those that depend on web traffic or online interest/conversions to generate revenue,” adds Kimberly Langers, chief operations officer at Rastegar Property, a private real estate firm in Austin. “For high-growth organizations especially, there needs to be an added focus on scale to ensure that leads generated from customers/investors don’t overwhelm internal marketing and/or sales teams, disrupting the user experience.”

Feeling overwhelmed is a common issue when it comes to statistical modeling, and it’s not hard to see why.

“Most advanced companies use statistical modeling in their activities to address different use cases,” says Jonathan Moran, global product marketing manager for SAS Customer Intelligence. “You need to know which model to use for which situation.”

Nonetheless, the appeal is also understandable. Companies are increasingly interested in statistical models for digital marketing because they help marketers to better understand their brands’ relationships with customers, says Tessa DeCandido, vice president of strategy at MediaCrossing, a digital advertising agency in Stamford, Conn. “In any business or industry, modeling is critical to driving business growth because it allows you to analyze a set of facts and then make a series of data-driven decisions based on those facts. The same is true in digital marketing. What audiences have the highest likelihood of becoming customers? What touchpoints were the most impactful in influencing audiences and encouraging path-to-purchase? Was certain messaging more effective?”

By capturing data across digital marketing efforts, companies can answer these questions, DeCandido adds. “Statistical modeling sounds daunting, but it’s really just a fancy way of saying: ‘We will help you analyze and understand your data so that you get smarter about your customers.’”

Ziech-Lopez agrees. “By investing in the right customer data up front, marketers are able to better understand how their brands and marketing communications are actually influencing their audiences. This approach does not have to be all-or-nothing. Every step you take toward being a data-forward marketing organization is a positive one, and no step is too small to potentially make a difference in your results.”

CREATING THE MODELS

The key to creating any model is first defining the questions you’re trying to answer, according to DeCandido. “What are you trying to learn? What problem are you trying to solve?”

This step is important, Moran points out, because the variables to include in the model will differ depending on the question the company wants to answer with the model.

Then, once you clearly define the outcome you are trying to achieve, “the next step is identifying the data at your fingertips to effectively begin to answer your question,” DeCandido adds.

Statistical models in marketing take on two basic forms:

• Audience models help companies understand their customers’ makeup. It includes factors like age, gender, income, where they live, who lives in their households, whether they have kids or pets, their shopping behaviors (do they prefer certain retailers, brands, or products?), and whether they’ve recently shown shopping intent, according to DeCandido.

• Campaign performance models enable marketers to identify the most effective strategies and tactics for connecting with and converting customers. They look at data like the media channels that drove the most engagement and conversions, the best performing formats, or the audience targets that were most likely to convert, DeCandido says. “Campaign performance models are critical for evaluating which combination of variables were the most impactful in achieving the desired outcome and then optimizing future initiatives to drive improved performance.”