Predictive Analytics and the Digital Marketer

What is predictive analytics? What is predictive marketing? Do I need a statistician on my marketing staff? These questions and others related to analysis and digital marketing are increasing in frequency. Since it is relatively easy to collect data, especially in the online space, digital marketers have a wealth of data points to describe their online customers and site performance. Wouldn’t it be great if we could reduce the data to its most relevant points and use it to not only describe, but prescribe and perhaps predict? Enter predictive analytics.

What is Predictive Analytics?

Predictive analytics is the practice of using data mining and statistical modeling to describe what is happening in your organization and estimate potential outcomes. As a result of predictive analytics, organizations are able to forecast revenue, define attribution models, and identify customer segments and score them on their likelihood to complete a desired action.

Predictive analytics is not new. Organizations have been using statistical modeling and data mining for years. It is used in fraud detection, revenue modeling, and even in HR hiring models. It is also used in direct mail campaigns – identifying households most likely to respond to the marketing message. The discipline is usually found in the Business Intelligence or IT groups within an organization, but is making itself known in the digital marketing space.

What is Predictive Marketing?

Predictive marketing is the practice of applying data mining and statistical modeling to optimize marketing efforts. For example, segmenting and scoring customers and online visitors based on a propensity to complete a desired action or defining attribution models based on campaign success to assist with budgeting and planning.

The challenge with digital marketing is the need to respond quickly. Most organizations have a BI or IT team churning through data and creating models to describe behaviors and outcomes, but the results may come after the opportunity. The web moves fast and digital marketers need to be able to respond just as fast. Having predictive capabilities within the digital marketing group enables marketers to not only analyze data quicker, but smarter. The result: improving the efficiency to act.

For example, it is the end of the quarter and online product sales are low. A quick statistical analysis takes in previous sales activity and behavioral data and identifies visitor groups and behaviors that are most likely to purchase given products. Then a targeting campaign with specific product and messaging to these groups is launched to provide more relevant promotional data rather than a generic campaign.

Do I need a Statistician?

Remember that high school or college math class? The one where you spent hours working on a homework assignment, only to come in the following day and have the teacher say “Now we are going to learn how to do it by computer”. The ability to calculate the standard deviation of a distribution by hand does have its benefits. But when it is crunch time and decisions need to be made, I am going to turn to the computer. Make sure you have the right resource for the job.

Having someone on the staff who understands the nature of the data and statistics will help. It does not need to be a statistician, but someone with the ability to analyze large data sets and perform statistical and econometric analysis. A plus would be someone with those skills and who understands how the business works, but those people are rare. I suggest having more than one person to bring various skills together, or supplement with consulting.

In Summary

Many people think predictive analytics is a black box, or crystal ball. Really it is just math and science. It is important to differentiate that it is not a crystal ball telling you the future and what to do next. It is compass pointing you in the most likely direction. The accountability to act is still up to you. Unlike a crystal ball or black box, you can take apart a compass to understand how it works. Similarly, you can break a statistical model into smaller components to understand the math and assumptions driving the outcome, which will help inform you of the decisions to make and the actions to take. The idea is to take the guesswork and subjectivity out of the decisions that need to be made and become a little more educated and efficient with action.