Google Trends, Ad Planner, and Site Affinity

The Advertising Research Foundation’s (ARF’s) Audience Measurement 3.0 Conference this week may be best remembered as the conference where Google announced its Ad Planner, inadvertently one day early (due to an agenda being posted on the ARF site prior to the press announcement or demo given by Wayne Lin, Google’s business product manager).Once I’ve had time to fully explore the Ad Planner and how it stacks up to Quantcast and other options, I may decide to report fully on the Ad Planner’s functionality, because there are clear intersections between search and display (which was actually another topic at the ARF conference).

Interestingly, the Ad Planner leak got agency conspiracy theorists in an uproar, with many claiming that this is proof that Google wants to make media planners obsolete. (Note to CMOs: if all your media planner does is look up site demographics to pull together a plan, you need to fire your media planner and agency.)

The Ad Planner is clearly more of a tool to streamline the planning process so that media can actually be tested in a live campaign and either be included more regularly (if metrics are good) or excluded from future campaigns. I never get bowled over by these demographically driven planning tools. In general, demographics aren’t nearly as predictive as behaviors, but they’re better than nothing. I actually prefer affinity scores (a listing of other sites the audience has also visited) that eliminate both the prior and the subsequent site visit. Google displays affinity in the media planner, but since you can’t get into the planner without a login, here’s an example displayed by its sister product, Google Trends, for eBay. Quantcast also has some free data that goes beyond Google Trends into the realm of the Google Ad Planner product. For example, here’s a similar result for eBay.

For media planning, one can make arguments both ways with regard to eliminating the pre- and post-visit sites from an affinity score. Clearly, if an advertiser is buying heavily on a site, the affinity score for that publisher compared to the advertiser domain will be unnaturally high. This makes the data useful for competitive research but presumes your competition knows what they are doing. I’m far more interested in knowing a pure affinity score that tells me where I can find audiences very similar to those of an advertiser (in the case of using the advertiser’s own domain) or a competitor’s domain. It’s very difficult to get true affinity scores because banner CTRs (define) (and even text link ad CTRs) are so low that if an advertiser is advertising heavily enough (and even when immediate click effects are filtered out), the consumer is often responding via search or direct navigation to a brand or URL mentioned in display advertising.

Google makes no specific mention of where the data used within either Ad Planner or Google Trends comes from, but I’ve found evidence that it’s highly likely the data comes from the toolbar or from a Google login cookie. It could potentially come from ISPs, but the number of instances that would be explainable purely based on ISP data is very low.

Let me illustrate. Within an SEM (define) or other online agency, there’s a high affinity between the SEM domain (perhaps due to employees reading or contributing to a blog or general visits to their employer site) and client-site visits. Agencies are often checking landing pages, researching keywords, or testing campaigns. In addition, agency teams are often looking at the competitor sites to check for offers or do competitive research. Both show up very heavily in both Google Ad Planner and Google Trends. I’d show you some links within the Ad Planner product, but most of you probably don’t have logins yet. However, I can make my point with a few links to the Google Trends site based on industry players of whom you are likely aware. I’ll leave it to you to guess which are the clients of the agencies/providers:

Clearly the Google cookie or Toolbar is the most likely source of this data.

Yet there’s no data for DoubleClick. Curious… And Performics shows industry links, so external users may be drowning out the internal people.

It’s also curious that given the popularity of Facebook, LinkedIn, and MySpace, that none of those sites seemed to have a high affinity. Perhaps the indexing process is about unusually high indices, which would eliminate those sites that everyone liked. If that’s the case, one might miss media opportunities within channels at these larger media properties.

All the engines are working to empower agencies with tools, but I’m not completely sold on the Google Ad Planner as a tool to be used in a vacuum yet. For it to be truly effective, it must be combined with other data sources. Most important, media campaigns must be actively managed.

Another ARF session that was somewhat interesting was one that covered the interaction effects between media and search. Dave Cavander, EVP, MarketShare Partners, and Joel Brodsky, VP, advanced analytics, Charles Schwab & Co., presented data that once gain validates the interaction effect between media (online and off-) and search. They proposed an elasticity-based media mix model to efficiently model overall spending and the mix to maximize profitable customer acquisition. If I get a hold of the deck, I’ll be able to include it in more detail when I cover this topic in the future.

Related reading

Screenshot shows a Google search for outdoor grills, the shopping ads shows images with “in store” showing the product is available nearby.