Analyzing the Google Content Network

The Google Content Network (GCN) can work for advertisers, though not nearly as well as a recent Google publication would have you believe. It takes a lot of work for most advertisers to get the GCN to deliver the measured ROI (define) that marketers have been accustomed to in their search campaigns. It's particularly challenging for online retailers to demonstrate the same ROI from content/contextual advertising they see from their search campaigns every day.

Google recently released a white paper, "CPA Performance Trends on the Google Content Network." If you were to read the paper, you might be convinced that the GCN was a performance powerhouse for nearly every advertiser. In reality, the study's methodology demonstrates that the GCN indeed performed well for some CPA (define) advertisers.

However, most online retailers are not managing campaigns to a CPA, but to a return on advertising spending (ROAS) or some flavor of ROI (immediate return or long-term lifetime customer value). Many, if not most, CPA-focused advertisers, on the other hand, want to generate leads or registrations.

The two consumer behaviors -- purchase and registration -- are quite different and represent two distinct stages of the buying cycle. Purchasers are clearly at the end of the purchase cycle and have made up their minds. However, while some registrants may be very close to the final purchase decision, most are still evaluating alternatives.

The Google white paper looks at more than 25,000 global accounts running both search and content (AdWords and AdSense). Half (51.6 percent) of the accounts had a CPA equal to or lower than their search network CPA. (Of course, this means that nearly half had CPAs that were worse.)

Even if you manage to a CPA instead of an ROI or ROAS, don't get too excited about revving up your AdSense campaigns in a bid to be one of the lucky 50 percent who have managed to get their accounts to hit their CPA goals. There are several things you should consider:

• Selection bias. If advertisers can hit their CPAs, they're more likely to keep running AdSense (keep content targeting turned on). Conversely, advertisers unable to hit their CPA targets would be expected to either settle for a lower CPA or turn off AdSense altogether. So, the advertisers from which Google collected data were less likely to be ones for whom AdSense didn't work. In addition, Google used only advertisers that were tracking with the engine's free conversion tracking tool (it's not stated, but I assume it's the AdWords tracking code, not the Google Analytics one). This is likely a selection bias because those advertisers using Google tracking may not be representative of the overall advertiser base.
  
  
• Nonstandard campaign settings. Those advertisers meeting their CPA goals tended to use either Google's Conversion Optimizer or site exclusion. Both of these settings improve the click relevance and conversion rate when properly deployed.
  
  
• Smart Pricing and lower CPCs. In addition to finding that content clicks were priced on average 28 percent less than search clicks, lower click prices mean you can reach your CPA target at a lower conversion rate. It's unclear whether the reduction in click prices reported in the research was the result of advertisers bidding separately (and lower) for content clicks or of Google's Smart Pricing algorithm contributing the majority of the average reduction in CPC. Smart Pricing is designed to normalize the advertiser's cost to make up for the lower likelihood of conversion to lead or sale.

My team and I have gotten AdSense to work for many lead-gen (CPA) clients, and even some e-commerce clients. In addition to site exclusion, we have found other techniques that help bring the CPA or ROI into compliance on the content side of the campaign:

• Campaign structure. Set up content-only campaigns and consider trying different ad creative, as surfers and searchers may not respond to the same messaging. Optimize landing pages for search and content.
  
  
• Use of site/placement targeting. Instead of simply excluding poorly performing sites (which show up in reporting only if they are large enough to have a statistically valid block of bad clicks), isolate good performers and potentially run campaigns against those exclusively.
  
  
• Day-parting. Measure the conversion rate on content clicks separately, and use either the Google settings or an API (define) driven solution to ensure that you aren't overpaying for clicks based on the predicted conversion and ROI/CPA.
  
  
• Geo-segmentation. As with search, some geographies may over- or underperform.

Many of the best practices in getting content to "work" in Google hold true for Yahoo, Microsoft, and the second-tier contextual players. Don't count content out of a direct-response lead-generation media plan or even an e-commerce plan, but also don't believe all the positive hype. Of course, if branding is also an objective, content advertising makes even more sense because your message get lots of "free" exposure, and you only pay for the clicks.

Meet Kevin at Online Marketing Summit in Boston on May 5, 2009.