How to Drive Revenue With Customer Response Modeling
A few years ago I worked on a marketing campaign for a targeted billboard campaign that had the whole marketing team tripping over each other. The campaign had a short turnaround time and we couldn’t agree about what was important. The creative person thought that the image was the only thing that mattered. The brand marketing person thought that the copy was the most important. I thought that placement was king.
In truth, all of the pieces put together are important, but I still firmly believe that placement is the most important. You could have a gorgeous ad with perfect copy, but if you don’t put it in front of people who care, none of that matters.
The same is true for direct marketing. A beautiful mail piece is important, but the success of the mailing depends on you targeting the right person. If you put a well presented pet supply catalog in front of someone who hates dogs and is allergic to cats, chances are that person will never buy and you just wasted your marketing budget.
Targeting the Right Prospects: What are Response Models?
Response models use data mining to find similarities between responders from previous marketing campaigns to predict who is likely or not likely to respond to a future campaign. The model is then scored against the prospects of the new campaign and a marketer can choose to mail only those people that are most likely to purchase. This increases conversions and decreases costs by only mailing to those most likely to respond.
Direct Marketing Models: Good, Better, Best
Not all models are created equal. Here’s a quick summary of different types of direct marketing models:
GOOD. Recency, Frequency, Monetary (RFM) models:
simple, better than not modeling
Though very basic, many marketers still rely on RFM models. Technically RFM models aren’t actually response models since they are descriptive but not truly predictive. This method emphasizes customer behavior and segments by how recently a customer purchased, how often they purchase, and how much they spend.
RFM can identify good customers and provide a lift in response rates versus not using any targeting. Another benefit is that it is both simple and descriptive, so it is easily understood by business people.
Unfortunately, RFM doesn’t take into account life stage and assumes that customers are likely to continue to respond the same way. If RFM is the only targeting method, the most attractive segments are likely to be over-marketed to at the expense of other segments that could be invested in.
BETTER. Traditional Response or Regression Models:
more sophisticated and predictive than RFM
Regression models determine the correlation between variables. Unlike RFM models, regression takes into account that scores can quickly change when combined with other variables.
The model is developed specifically to predict a desired behavior, such as response. Response models require both responder and non-responder data to identify patterns of those likely to respond to a marketing campaign.
This is by far the most widely used approach for marketers and has been a mainstay of predictive analytics for decades.
BEST. Multi-Channel Customer Level Response Models:
A New Approach that Outperforms Traditional
This innovative approach identifies not only those prospects most likely to purchase, but also which marketing channel they are most likely to respond to. This allows marketers to optimize their marketing budgets most effectively by contacting the prospect in the channel(s) they prefer and are most likely to be moved by.
Multi-Channel Customer Level Response Models are different from traditional response models in that all of a prospect’s known activity is taken into account – email opens, web browsing, display ad click-throughs, mobile, purchase behavior – and not just direct mail behavior. With a more holistic view of the customer, a marketer can create the ideal customer contact strategy for each customer.
In a recent head-to-head in the mail test with a major cataloger, the Multi-Channel Customer Level Response Model outperformed the traditional response model with a more than +14% lift in response rate. This increase translates into millions of dollars in new annual revenue at the current spend. A new subject line or pretty picture won’t drive revenue like that!