Increasing Your Prospect Response Rates by 34%
News Direct – Number 53 | May 2006
Increasing Your Prospect Response Rates by 34%
by Arthur Middleton Hughes, KnowledgeBase Marketing, Inc.
Some experienced insurance mailers think that they have all the answers. A lot of them, however, are not using prospect databases for their acquisition mailing. If you are one of these mailers, you will be interested in learning how some companies have been able to increase their acquisition response rates by more than 30% while reducing their mailing costs by almost 10%. If I have gotten your attention, read on.
Today, all insurance companies are actively renting lists of likely responders. They send them direct mail offers, hoping for a nibble. A typical mailer may rent 300 lists totaling several million names. After the mailing, he is required to erase these names from his mailing files, keeping only the one percent or less who respond. What does he learn from this process? He learns what lists work best for him, and which offer gets the best response.
What he does not learn from this mailing is some vital information: which age group, income group, and ethnic group responds best. He would like to know who responds by housing type, family composition, lifestyle, and about twenty other key factors that we all know are important in finding the right insurance customer. Using prospect databases, some insurance companies today are able to select their prospects by these very factors, instead of what list they came from.
The way a prospect database works is this. You rent names for an entire year, instead of for a single mailing. You keep the names in a database, and pay the owner every time one of his names is used. Because you have the names for a year, you can afford to append all sorts of demographic and behavioral information to the records, which you certainly could not afford to do if you rented the names for a single mailing. In addition, you can keep the promotion history in your database. Every time you mail a promotion to someone, you put that fact, the promotion used, and the date in the person's database record.
What kind of data can you keep on each prospect in your database?
With this information, you can build models that tell you which of the factors had the most influence in determining those who responded and those who did not respond. A good model looks at each of these appended factors and assigns it a weight (either plus or minus) which indicates how important it was in determining the final outcome (the person responded and bought a policy, or did not do these things).
Here are the actual results of an auto insurance company mailing results, before and after he created his prospect database:
As you can see, with the model he mailed exactly the same number of pieces, but his response rate, sales rate, revenue, and profits went way up. The only thing that went down was his revenue per sale.
Further Results from a Prospect Database
Step One:Appended Data. Most rented response names come to you with only the name and address. You do not know the age, income, length of residence, or any of the many factors listed above which are used by a model to select the likely responders. To get this additional information, you have to have a service bureau, like KnowledgeBase Marketing or Acxiom, append this data to your rented names. This is not cheap. It will raise the cost of your mailing. There is a way around this problem.
Step Two:Compiled Names. If you rent compiled names such as AmeriLINK' from KnowledgeBase Marketing these records already have all the data you need already appended to them. What's more, compiled names are about half the cost of the response names you have been renting. There is normally a catch in this, however. Half of the households in America never respond to direct mail. So, normally, compiled names (which are every consumer in America) usually get lower response rates than response names (which are people who have responded to direct mail and bought something).
Step Three:Using a Model. This is where the model comes in. If you have a prospect database and have already done some significant mailing, you have the results: for example, you have four million people did not respond and buy and 40,000 who did respond and buy. A good model will use the data from these previous promotions to accurately predict who will respond to a future mailing. It will arrange any list of U.S. consumers in deciles based on their likelihood of responding. The model will give you a chart that looks like this:
Step Four:Scoring the Names in Your Database. The end product of the model is an algorithm (a mathematical formula) which can be used to score a database of names to determine which decile each name falls into.
Step Five:Using Both Compiled and Response Names. A model has a second advantage which is really profitable. The model will work with both response names and compiled names! That means that you can build your prospect database for about half the cost of what you are paying today. Here is a chart showing before and after costs of an illustrative insurance mailer with a prospect database. He shifted from using mostly response names to using mostly compiled names for his mailing:
As you see, he is mailing exactly the same number of pieces, 13.6 million, but his cost of renting names is about one third of what he had been paying before. Since compiled names already have the data appended, he had to pay for data appending only for his rented response names and for his house file names.
Step Six: Mailing to Segments. You would not send the same mailing to a 65 year old prospect that you would to a college student. You would not send the same mailing to a couple with young children that you would to empty nesters. And you would not send the same mailing to someone with an income over $100K that you would to someone with an income of under $30K. To do your acquisition profitably, you should create segments of similar prospects, and send a mailing piece that was designed specifically for that segment. Renting response names, you usually do not have this kind of information, so you really cannot do much segmentation. With a prospect database, loaded with relevant data, every mailing can use segment-based strategy, with far higher response rates and sales.
Step Seven:Modeling on Conversion Rather than Response. It is all very well to get people to respond to your mailing, but the real payoff is if they buy a policy. What good are a lot of responders who do not buy? The solution to this problem is to base your model on those who actually purchase a policy rather than those who just call or go to your website. There can be quite a difference. If you have a prospect database, you can build two models: a response model and a conversion model. On your next mailing, test both models to see which is giving you the most bang for your buck.
A Recent Case Study
There are ways around all of these problems.
Should You Build a Prospect Database?
Editorial contact: Linda Barba, Project Marketing, Inc., 610-889-2036, email@example.com, (1,961 words).