Context Is King, or Is It?

From the early days of search engine marketing (SEM), content has been king, both for organic search engine optimization (SEO) and pay-per-click (PPC) search (keeping PPC editorial reviewers happy). Now it seems context is king in the world of SEM… or is it?

Recently both Google and Overture have announced expansion of their networks to include contextual advertising. About’s Sprinks has been serving up contextual content with ContentSprinks for a while now, and the other smaller PPC engines have been supplementing search inventory with contextual inventory for years.

Clients have been calling and emailing me to ask if this contextual inventory is a good thing. The questions have run the gamut: What exactly is all this contextual inventory? How does this inventory differ from the search engine results inventory we love, but hate to pay for? Will this contextual inventory affect us?

Essentially, the contextual inventory attempts to match user interests to the ads being served through contextual relevance. Some examples of contextual relevance are:

  • Editorial content about a specific topic is matched with in-page, text-based advertising units that look much like the search listings we see in portals. The contextual ad engine boils down the essence of the page to a search term and serves up ads marketers have purchased for that term. The number of text listings varies by implementation.
  • Editorial content results in a contextually relevant ad unit being served in a pop-up or -under. The pop-up/-under typically contains five listings pulled from the search engine listings database.
  • Registered domains not in use (parked) are covered in a set of listings in directory format. Some of these listings relate to words used as part of the domain name; others are just popular search terms organized in a directory format.
  • You have a “helper application” installed (less flatteringly called adware, spyware, or scumware). Based on the sites you are visiting or the actions you are taking, the helper application launches a window filled with text links from a PPC network. Helper application companies include Gator, eZula’s TopText, and WhenU.

Effectively implementing contextual inventory placement hinges on a really accurate mapping technology. The technology extracts the “true essence” of a page or site, then finds the phrase that most accurately describes that essence. We hope for a truly impartial contextual algorithm that will not include price paid for clicks in its mapping decision.

Google’s Algorithms Search for True Essence

Google’s new Content-Targeted AdWords are based on the first example above — ads placed within a publisher’s page — and are being syndicated across evermore content sites, including HowStuffWorks, Weather Underground, and Blogger (surf around on HowStuffWorks for examples of the ads). According to Google, the company’s famously accurate relevancy algorithms determine the best matching keywords and key phrases based on what the Google index understands about a page.

Google is providing AdWords advertisers with free contextual traffic until March 12, 2003, at which point the contextual network is an opt-in option. My firm has been running and testing Google’s Content-Targeted AdWords for several months as part of the beta test of the program, and, generally, we agree with Google’s assertion the return on investment (ROI) on contextual ads is “on par with ads shown on search pages.”

We saw conversion behavior (percent conversion) from the contextual ads that in some cases exceeded search by 15 percent. However, for some client campaigns, the conversion rate on the contextual portion of the campaign was significantly below pure search (an over 30 percent drop in conversion).

So keep a close watch on your campaign conversion with the overall campaigns you have running. If your campaign conversion drops with the introduction of contextual search, you can always opt out of the contextual network. I hope Google will make it easier to identify the contextual traffic through use of a special referrer tag. That will make it easier for marketers and their agencies to make informed decisions about contextual inventory.

Overture Experiments With Contextual Ads

Overture recently announced it will be entering the contextual ad market. In reality, Overture has been experimenting with contextual inventory for quite some time, on several fronts.

Applied Semantics is an Overture network affiliate, and its inventory is contextual. Applied Semantics also serves inventory, based on contextual relevance. Furthermore, Overture is testing Gator’s new method of generating contextual pop-ups, Search Scout. Searchers perform a portal search, and Gator’s Search Scout spawns related listings in a pop-up ad unit, effectively extending Overture’s reach beyond the Overture network. So, if a searcher has Gator installed, a Google search may result in a Gator pop-up containing Overture listings. It is not clear exactly when Search Scout listings are triggered (i.e., when the pop-up occurs). Is it immediate or delayed for a certain period after the searcher has reached a destination site?

Thus far, Overture has not allowed its advertisers to opt out of the contextual portions of its inventory. Perhaps, based on the company’s recent announcements, its plans include a separate network that provides the marketer with opt-in control or even separate pricing. Time will tell.

We recently conducted an analysis of a large volume of Overture traffic across many clients and based the analysis on original traffic source (referrer). The results were enlightening, and for some clients the results were similar to the Google contextual traffic test. Some of the traffic coming from the contextual Overture network converted well; other clients saw significant drops in conversion behavior from the segment that originated from contextual links.

Great Promise but Increased Measurement Need

Why the diversity in contextual versus search conversion changes between clients? Well, contextual advertising quality relies on the ability of the contextual engine to match the keyword to the content, the location of the contextual inventory on the page (contextual proximity), and the quality of the original content to which the ad inventory is being matched. In addition, some industry segments may lend themselves better to contextual advertising than others. This would result in some segments of a contextual network being better suited to certain industries and types of ads.

Based on the results of testing on the top PPCs and several second-tier PPCs, I believe contextual inventory shows great promise. However, there is a diversity of quality in networks and contextual mapping methods. That means measurement and metrics are even more important.

Contextual inventory is here, and chances are you are paying for some of it now. Make sure you know what you’re paying for and that it works for you.

Related reading