A Targeting Tutorial

Targeting. Whether it’s carefully selecting sites to reach a precise demographic, or using sophisticated profiling techniques to find that needle-in-the-haystack customer among millions, everyone knows the value of this powerful one-to-one potential of the web.

First, what does targeting mean? Targeting can be defined as reducing the pool of available viewers who are going to see your advertisements. Unless you have tremendous resources at your disposal, you’re probably never going to be able to show your ads to everyone using the Internet. Thus, by definition, every media buy is targeted to some extent against a subpopulation of the Internet. We all target! So there’s nothing inherently magical about targeting.

The web’s ability to display ads only to people living in the 94105 zip code, who use a Netscape browser, and who have purchased something on your e-commerce site in the past, is both fascinating and satisfying to marketers. Such fine targeting seems to fulfill the web’s promise of one-to-one marketing. Advertisers who were weaned on traditional media targeting are astounded by tales of web technologies that filter through mountains of data and millions of people to find those that are just right for you.

Make Sense Of All Your Options

Because of the popularity of targeting on the web, it has captured the attention of many technology providers. Networks, ad servers, and large web sites all promise to offer different and unique targeting options that leverage some web technology to focus the audience that the buy will reach. Sometimes targeting options are based more on what’s technically possible than what’s useful in driving results.

Let’s say you’ve decided you’re going to run a targeted buy. You approach a large web site or advertising network and ask about the types of available targeting options. Chances are you’ll probably get some subset of this menu:

Targeting by Content Run your campaign on specific web sites, groups of web sites, or subsections within web sites that attract an audience interested in your message.
Targeting by Browser or Operating System Show your campaign to web surfers with a specific computer configuration.
Targeting by Time Show your campaign to web surfers only at specific times of day or days of week
Targeting by Organization Name Show your campaign to web surfers who are affiliated with particular companies or organizations.
Targeting by Geographic Location of ISP Identify the physical location of a web surfer’s ISP and run a campaign only against users from a particular geographic area.
Targeting by Profile Data Show your banner only to people who have volunteered specific information about themselves that is of interest to you (for example, their gender or hobbies).
Targeting by Inferential Data Identify those groups of web surfers who would be most responsive to your message based on inferential preference data gathered about them as they surf anonymously around a network of sites or a large web site.

With all these menu choices, how do you know which one to choose? A particular technique might work well for one advertiser, or campaign, but not at all for another. As an return on investment (ROI)-oriented advertiser, how do you sort through these options to find the technique that will work best for you?

Make Targeting Work For You

The best approach for ROI-oriented advertisers might be to test and evaluate. Don’t try to guess up front whether web surfers from 94105 or 02912 are going to respond better. Let the behavior of your potential clients tell you what to do.

  • First: Run a campaign across a large network or several large sites. Resist the temptation to “optimize” response after a few days. Instead, wait until you’ve generated a good sample of response data. This becomes your baseline response rate. I recommend allocating 20 percent of your budget to determining your baseline response rate.
  • Second: Examine the response across multiple targeting vehicles. Look at which web sites (or sections within sites) generated the best response. Consider which times of day and days of the week might have produced more clicks or transactions. Think beyond the obvious. Look for reporting tools that quickly show you which web sites, banners, times of day, days of week, browsers, operating systems, zip codes, U.S. states, and countries performed the best for you.
  • Third: If the extra response you’ll get from these high-performing vehicles is worth the extra cost of targeting against them, skim off those vehicles and heavy up against them. If you’re spending $50 cost per thousand (CPM) on a targeted buy, you should be getting well more than five times the response that you would be getting on an untargeted vehicle at $10 CPM. If the extra targeting is not worth the extra cost, spend the remaining 80 percent of your campaign in untargeted vehicles.

Once you find your high-performing targeting vehicles this way, a large network or site can usually deliver much more of that target audience, thus enabling you to “clone” your high-performing audiences.

After developing your baseline and successful targeting plan, you may notice that your target audience(s) stops responding to your message and any fresh new creative you try. Your high-performing target is simply burned out. It’s important to keep running new messages against a broad and growing audience, to build your next group of high-performers in the network. A large site or advertising network with broad reach gives you access to new eyeballs all the time.

Given the assortment of targeting options available and the ease and low risk of running a test on the web, testing is the only approach that makes sense. You’ll save yourself the effort of racking your brain about all the targeting options that might work best for you. Instead of targeting for targeting’s sake, you’ll rely on actual results to tell you which targeting options work for you.

Related reading