Keep Testing, Changing Your Web Site
For example, while on a panel at a recent online advertising conference, I polled the audience. "How many of you run a Web site where you attempt to get registrations or sell something?" Every one of the 200 people in the room raised his hand.
Next question: "How many of you are aggressively testing the parts of your site that demand a response -- by that I mean, how many of you have tested at least 10 new pieces of creative on your site in the last two weeks?" Five people raised their hands. Less than 3 percent.
A Web site is not a one-shot event. It's a process. And the only way to know if a site is doing well is to measure. And no doubt, measurement will reveal disappointing acquisition and retention costs. After all, the numbers are rarely as good as we would like.
What to do? Test.
Sure, testing is not a revolutionary concept, but it works 50 times better online. At Yoyodyne, testing has increased conversion rates by a factor of 20. That's a 95 percent reduction in costs.
On the other hand, notice that the home pages on Amazon.com, Yahoo and ZDNet always look about the same. The reason: They're not testing very well.
The Internet makes it possible to test 100 pieces of creative in an afternoon. Web marketers can test color, size, wording, offer, arrangement, copy, proximity and any other variable imaginable -- for free -- in an afternoon.
Why not put 2 percent of your Web-site visitors through one of a series of alternatives designed to beat the control?
Why not test the congruence between banners and actions? When a banner's creative approach matches the offer, results go through the roof. Then the challenge is to test it against a higher click-through banner and see which one gives the best overall performance.
All the rules of traditional direct marketing apply. Get a good sample. Don't test more than one thing at a time. But most importantly -- measure.
If a site can't differentiate the conversion rates between two banners or two sources of traffic, then why bother testing creative?
Knowing which has better click-through is worthless without measuring conversion. Not convinced? Try running a banner that says "Click for Naked Photos of Queen Elizabeth" and watch the click-through rate go through the roof. But odds are that there won't be much of a bump in sales or acquisition efforts.
Testing lets marketers pay media directors less, too. Instead of relying on media directors for impassioned, emotional decisions involving editorial environment and synergy, just test everything and let the market decide.
Suddenly, it's understandable why pay-per-click deals are so attractive to many direct marketers online. Measurement allows marketers to buy Web-site traffic, instead of just focusing on the creative elements of banners.
Here's some simple math: What percentage of visitors to a site perform the desired action? What is that action worth to the company? Find advertising that delivers visitors at a lower cost than the desired action is worth and buy as much as possible.
However, don't focus too closely on bringing down the cost of traffic. It's better to boost yield instead. A small increase in the last step of the process has far more effect on the bottom line than a small decrease in traffic costs.
Though the online advertising debate is currently focused on targeting and technology, Internet marketers who test won't have to be wowed by cool shockwave demos. Instead, they'll be wowed by sales figures.
Seth Godin is president of online sweepstakes marketer Yoyodyne, Irvington, NY. His e-mail address is firstname.lastname@example.org.