Remember when pretty good targeting was enough? Internet ad serving technologies now enable micro-targeting by industry-vertical-and-job-function-and-relevant-niche-content, and it’s hard to find anyone who’s satisfied with plain old targeting anymore.
When he’s marketing information management products that help developers do their jobs better, Mike Paradiso, CA’s VP of Worldwide Media, sees better click-through rates from banners run on developer-only sites than those same banners running on broader business or IT sites. And Kim Kochaver, Management Supervisor at MRM Gould, has had similar experiences: niche targeting generally drives higher click-through, if not always lower effective cost-per-click rates given the premiums charged for such narrow rotations.
At the same time, though, both rattled off a list of exceptions and caveats: ad creative may have more impact on response rates than targeted placements; micro-targeting for one audience attribute (say demographic or job function) ignores other attributes (like psychographic mind-set); and “sometimes a less precise placement will surprise you with terrific response.”
The long list of caveats calls into question the simple logic that the more specific the target, the better.
Kim and Mike both talked about the perils of over-targeting, or “lazy” micro-targeting. Enterprise server ads that run only adjacent to stories about servers, for example, might drive better-than-average response (not always though, see below). But it’s sloppy marketing that misses an opportunity to message to a prospect who needs servers to solve a related technology challenge. Fixing lax security, upgrading operating systems, and revamping IT management strategies will, in all likelihood, require new server hardware; yet careless micro-targeting ignores these prospects.
Looking at click-through data as a proxy for interest, we begin to calculate the real (if partial) cost of over-targeting. Vendor participants in ZDNet’s “PowerCenter” sponsorship program provide educational assets (whitepapers, case studies, webcasts, etc.) that ZDNet integrates as text links on most pages of the site. Like pages of the ZDNet site, all vendor assets are assigned to categories — hardware, software, networking, security, etc. — to facilitate category matching, one flavor of micro-targeting. When the ZDNet producers override the category-matching system, however, human insight frequently out-performs the automated ad-delivery system. The best-performing text links in April 2004 promoted on related pages (0.6% CTR). The next-best performers were operating-system evaluation kits posted on related pages (0.5% CTR) and hardware-related pages (0.4% CTR). Operating system software assets, in fact, performed better on server hardware pages than server hardware assets on those same pages.
If ZDNet’s enterprise security content turns a million pages a week, 0.6% is 6000 prospective customers eager for buying information on that vendor’s server products — steep opportunity cost in the name of marketing efficiency. And that doesn’t count the 99.4% of pages that didn’t drive a click.
Prospective customers don’t always structure their research according to the taxonomy of websites, nor do they self-select into database marketing buckets. As a result, job-function filtering faces the same over-targeting risk as content matching: According to IntelliQuest’s CIMS Business Study 2004, only 19% of those who’s primary job responsibility is writing software programs actually report their primary job function — essentially their title — as application development or programming. I’m guessing this phenomenon affects other job functions as well.
Overly eager micro-targeting promises efficiency and often delivers it, just like turning off the lights saves on the electricity bill. But turning off the lights also comes at a cost — your staff isn’t productive, isn’t closing sales, and isn’t manufacturing new product when the lights are out. Over-targeting can limit wasted impressions (though given the evidence, let’s be careful what impressions we call “wasted”), but it also abandons many prospective buyers who don’t fit neatly into the boxes we’d like to put them in.