Ammon Brown

This article was originally published on Adotas.com.

When running a large-scale search engine marketing (SEM) campaign, there are some things that marketers typically assume. For instance, it’s natural to assume that search ads appear in search and content ads appear in content. Constantly policing this would be far too time-consuming to be worthwhile. That assumption, however, appears to be slightly flawed.

Recently, while reviewing some of the search query reports for one of our clients, we came across some very interesting trends. The query with the third-highest impression volume was “prints patterns pillows pillows throws home décor home.”

This query triggered the keyword “decorative pillows,” which was set to Google’s broad match. However, this was very obviously not the kind of query that a user would enter in a search box, and certainly not 20,000 times, as the impressions in the report would suggest. This query also sported a click-through rate (CTR) of .02%, very similar to what would be expected on the content network.

To find answers, I did what anyone in our industry would do… I searched. I copied the long query and pasted it into Google. The top result was a page on Target.com that bore a page title similar to the query I saw. When I looked at the bottom of the page on this site, I saw my client’s ad. This client is opted out of the content network, so how could their ad appear in a context such as this?

I dug a little further and looked at the source code of Target’s site. I found that their site was using Javascript to send its page titles to Google as a search query through the AdSense for Search program, not through the content network. In essence, Target is showing search ads on a content page using a script to simulate a user-entered query.

When I went back to my search query reports, I found over a thousand queries that bore similar characteristics to the one I initially investigated. Spot-checking these by searching them on Google invariably led me to pages on Target.com. When I aggregated the performance of these queries, I found a cost per conversion that was more than triple that of the rest of the search partner network and four times that of Google.com search.

In addition, I discovered a few other sites, including Macys.com, that were sending queries using a similar method. However, many of the queries sent by scripts on these sites look very much like standard search terms, making them difficult to identify and evaluate separately from actual user-initiated searches.

The big issue here is that my client was paying search CPCs for a placement that is a content placement with no way to set a bid based upon performance. Since Google does not provide an option to block individual sites in the search partner network, nor do they allow us to set different bids on a site level or even for the partner network, I came up with a solution that appears to be working.

First, I duplicated all of my campaigns. All of the original campaigns were opted out of the search partner network while all of the duplicated campaigns were opted in, but their bids were lowered substantially. I also set many of the hardest-hit keywords to exact match and lowered the bids on any keywords that had a CTR that looked like a content placement.

The results were immediate. The search partner network now provides our client a cost per conversion close to the level we see on Google. Our aggregate CTR jumped from .5% to 2.9%. Functionally, we gave Google keywords with a higher CPC that they will select to show on a Google search before the lower-CPC duplicates, but we gave them only one option at a lower CPC when the query originates from the search partner network.

A great number of competitor retail sites were included in the ad units on the ecommerce sites I came across. I recommend that all search engine marketers look through search query reports, separate out the search partner network, and look for any queries with a very high impression volume and a very low CTR.

You should be especially concerned if that query looks like a website breadcrumb or is nonsensical. If you find that you are affected, you may want to consider duplicating your campaigns and taking some degree of control over these placements.

So the question remained, how can this be within Google’s AdSense terms of service? A little bit of further exploring led me to a page in Google’s help documents for advertisers that suggested that search ads may appear on pages within a site’s directory: “Your ads may appear alongside or above search results, as part of a results page as a user navigates through a site’s directory, or on other relevant search pages.”

Upon asking my Google rep, I was told that this is, in fact, an example of this kind of directory placement and some “premium partners” are allowed to do this. So basically, Google considers drilling down into the content of some sites to be a type of directory search, even though there is no query entered by a user.

What Google tells their publishers appears to be substantially different. Google’s terms and conditions page for AdSense for Search partners states, “Queries must originate from users inputting data directly into the search box and cannot be modified.”

This seems pretty clear-cut; AdSense for Search can only be initiated by user queries. Why, then, does Google make exceptions for “premium publishers?”

This seems incongruous, as it seems that Google disallows this practice in general, but provides exceptions for a few large retailers who subscribe to AdSense. It is possible that Google has some internal mechanism that evaluates a “directory” site against some unknown criteria for inclusion in the directory search category.

It is also possible that some premium publishers did not want the considerably cheaper content ads on their pages and lobbied for the higher-yield search ads.

The bottom line is that advertisers have zero transparency, zero control and usually zero knowledge of the existence of this practice. Regardless of why this is happening, Google will change this practice only when enough advertisers have made enough of a clamor to effect change.

If every advertiser reduced their bids to a level commensurate with the worth of this type of traffic, it would effectively bring the overall CPCs down to the worth of the lowest common denominator in the network, including the CPCs on Google’s own site. I doubt that is something Google wants.