One of the only true ways to find out if something in your direct-marketing promotions works or doesn't work is to test it. Like an experiment in a laboratory, you can test one controlled sample of mailings (that you know has worked in the past) against an experimental sample. In sending two versions of the same mailing out to separate segments of your list, it's possible to track response and sales and measure which version generated the most business.
In designing future mailings you will be able to use valuable data, which tells you which kinds of copy and design have worked best in the past. The most effective way to test copy and graphic design elements is to focus on very specific elements in isolation. This is true A versus B panel testing. An example would be testing two different headlines on a retail postcard mailing with a coupon. Version A could say "Come on Down to Our Grand Opening," and version B could say, "50% Off Select Merchandise." Aside from the element being tested, the A and B versions should be identical in terms of copy and graphic design. Avoid performing head-to-head tests of totally different types of promotions because there will be too many different variables to determine which elements pulled their weight. For example, sending out a direct-marketing postcard with one set of graphic design and copy against a letter in an envelope with totally different set of graphic design and copy, will not provide specific enough information about what works. Testing should tell marketers specific things like, "This headline worked," "We need to do four color," or "Customers responded more to the photograph of the entire family than the image of a mother and her child." When conducting a test, it's important to build the testing information into your database so you can later track which customers received which version, and how that affected sales. The more contacts you send the test to, the more reliable the results. Panels of 1,000 or more names in a mailing of 2,000 copies are preferable, while the gurus of testing will tell you that at least 5,000 names out of 10,000 total are necessary for the most accurate results.
No comments yet.