This question is best answered by example and by thinking about how things have started to change in the marketing world. Nowadays, businesses can measure and record many aspects of their interactions with individual customers. This information can be used to help businesses better understand which customers are most interested in their product, what their customers’ needs truly are, and which marketing efforts are working best for each customer segment. Once businesses better understand their customer-business interactions, they can make data-driven decisions to improve these interactions. In some cases, this data can even be used to make predictions about new customers.
Making predictions involves creating a mathematical or statistical model which learns from historical data to forecast future events. Being able to predict which customers will interact with your product allows businesses to prioritize their marketing efforts and marketing spend.
As an example, one of our clients is a lending company which keeps track of every customer which was both approved and not-approved after they’ve applied for funding (this company counts conversions as customers who were approved). In addition, the company also keeps track of each applicant’s credit score, industry, state, and annual sales information. Armed with this data, we leveraged statistical modeling techniques to create models that could predict the chances that new customers will be approved. In other words, using the model the company can now classify new customers as “fundable” or “not-fundable”. Why is this useful?
To see how the model could be used, suppose that the company will spend $100,000 over the next few months marketing to new customers, and that 40% of those new customers will not be approved for funding and 60% will be approved. These percentages are not known in advance because the company (obviously) cannot see into the future. This means that $100,000 x 40%=$40,000 will be wasted on non-fundable customers and $100,000 x 60%=$60,000 will be spent on customers who are funded. Alternatively, if the company had used the predictive model from the outset, they would have identified those non-fundable customers and specifically not spent money marketing to them. If the model was perfect, this means the company would have only spent $60,000 instead of $100,000 to acquire the same number of total conversions, thereby reducing their cost per conversion by 40%. In this example, the data collected and the predictive model saved the company $40,000 which can be used towards other marketing efforts or investments.
This is just one of many different applications of predictive marketing. In general, predictive modeling should be used to answer specific business questions such as: which markets should I decide to attack next? How should I adjust my budget over various channels to increase revenue? What products should I recommend to my existing customers? One important requirement for predictive modeling is to have high quality data. There are several ways to approach this:
- First define your business goals and then decide what data is best to record.
- First collect a bunch of data and then mine the data for insights.
- Augment you existing data with 3rd party data sources.
Most people would probably say that option (1) is best, but it all depends on your own specific goals and limitations. As newer data-recording and data-storage solutions are created, the demand and utility of predictive marketing will continue to climb. From this perspective, marketing is destined to become an increasingly scientific field. So, we will have physicists, biologists, chemists, mathematicians, and marketicians. Marketicians? Maybe marketologists is a better name. Or not, time will tell.
To check a sample predictive report, click here.
