One of the best ways to fill the early part of the online marketing funnel is to advertise on Google’s network of partner sites, the Google Display Network (GDN). Essentially, these are websites that are either showing AdSense ads or websites that show ads from DoubleClick (some of the GDN inventory is actually “remnant” inventory that no one made specific placement buys for on DoubleClick).
SEMCopilot, our AdWords management software for PPC account managers, collected data throughout Q1 from both beta testers and paying customers’ AdWords accounts; here is some analysis of their GDN placements.
How our data is organized
To make analysis of placements quick, SEMCopilot’s “Placement Insights” feature automatically sorts placements into 40+ categories. Google has its own categories you can use for exclusion purposes; we assume that most customers are already blocking porn or adult sites and so on. The purpose of our categories is to gain insights into where ads are running, for brand protection purposes and for performance analysis.
Based on our users data, the sites we’ve categorized represented 96 percent of the impressions they received on the Google Display Network in Q1.
Why bother categorizing sites?
Because of the number of sites on the GDN, simply examining a placement report is a pretty fruitless effort for most AdWords account managers. At best, they typically look at the top several hundred… then, when the number of clicks they are looking at gets down to one or two, they stop looking further. It’s a very long tail of sites.
Thus, organizing placement reports into categories makes it much easier to glean actionable insights. Below are actual quotes from users who have examined our Placement Insights Report — they’re illustrative of the surprises one often discovers:
“Hey, my ads are running on dating sites!”
“Hmm… what are all these political sites?”
“Whoa, why are my ads running on eBay?”
“What the heck are my ads doing on all these foreign-language sites?”
“Wow, [category] really gets a lot of conversions. Maybe I should do a placement campaign targeting those sites and bid up on them.”
The categories we examined
Most marketers can intuitively look at categories and rapidly decide on certain ones that won’t work for them. For instance, if you’re selling architectural design services, “Comic and Anime” is unlikely to be a high-performing category.
That said, for remarketing campaigns, you’re following your audience around, so who knows? Maybe they like those kinds of sites! The key is always to analyze your conversions and cost/conversion by category; then you can get a sense of where your advertising dollars should best be allocated:
Okay, where are the ‘fake news’ and ‘extremist’ categories?
You can’t read Search Engine Land or Advertising Age these days without seeing tons of articles about “brand protection,” “fake news,” or how Google is profiting from some ads on some horrible website or allowing ads to run on some pro-ISIS website, or whatever. Just for the record, we’re not in a position of judging what news is fake or who is extreme, and we eagerly await Google’s improvements in those areas. That said… well, there are news sites, and there are news sites.
We do have a “News” category which contains mainstream news sites. We also have a “Viral Amusement” category that contains the type of websites I’m sure you’re familiar with — they appear to be newsy, but articles have provocative photos, with headlines ending in “…and you’ll never believe what happened next!”– usually next to some AdSense ads about reducing belly fat (you know, those ads with the cartoon drawing of bananas).
Certainly some “fake news” sites may appear in our “Viral Amusement” category, but this is simply because we find that these types of sites don’t really perform as well as mainstream news sites do. And from a brand protection standpoint, often the pictures are NSFW and could result in your upper management asking you, “Why are our ads running on these sites?”
So, which categories performed best and worst in Q1?
Below are the four highest average cost-per-conversion categories across all SEMCopilot customers (top) and the four lowest (bottom).
One amazing anomaly that resulted in a very actionable idea
As we were examining the data, one thing really stuck out. Based on our categorization scheme, 96 percent of our customers’ impressions were covered. But the “Uncategorized” sites, while representing only 4 percent of impressions, represented 25 percent of spend, 35 percent of clicks, and a whopping 65 percent of conversions!
Our first theory was that this must be some sort of click fraud. Maybe there are sites out there on the GDN using bots that are automatically filling out forms, resulting in worthless leads, in the hope that you’ll placement target or bid up on them. We took a look, and although we did find four or five sketchy-looking sites with suspiciously high conversions, the volume wasn’t enough to influence the numbers much.
No, when we took the “Uncategorized” category and started categorizing the *converting* sites within it, we found that the vast majority (over 90 percent) were parked domains! Not only that, they were parked domains that were variations on the customer’s brand name!
As it turns out, parked domains are a GREAT place to advertise
For years, Google AdWords reps have been scolding me for leaving Parked Domains on in GDN campaigns. Clearly, they have been disabling them as a suggestion on some sort of checklist. I’ve always left them on because I suspect that people mistyping your brand name may navigate to your site.
Our data clearly bears this out; most of these suspicious-looking conversions were simply from parked domains that were either misspellings of or slight variations on the customer’s brand name. So they are really brand navigational queries. It may be irksome to pay a domain that is infringing, or maybe close to infringing on your trademark, but trust me, you want those customers.
Just don’t advertise on all of them
Of course, not all parked domains will perform well; only ones that are similar to your brand name seem to do the trick. I would suggest turning them on for a quarter and discovering which ones are working. Create a placement campaign for those, and bid way up on them — then turn off the AdWords “parked domains” category in your original campaign.
We roll out new features on a monthly basis, so rest assured I’ve just added a “Parked Domains” category into the hopper. This should make it really easy for our customers to identify likely brand variation domains that they should placement target.
The biggest actionable takeaways I see from the data are:
- YouTube may be worth testing placement targeting on, provided you layer on other targeting methods such as keyword targeting, due to its low CPA. Note that I don’t mean video campaigns, I mean simply running GDN text (now called “responsive” format) ads targeting YouTube.
- Political sites actually have the potential to perform great for you, provided you’re not too worried about brand protection implications. (By “political sites,” I am largely referring to political news sites — e.g., breitbart.com, huffingtonpost.com, and so forth.) It could be they simply are less popular with advertisers and thus have not been bid up as much as more mainstream news sites. In fact, with the recent campaign against breitbart.com, I would imagine you can pick up a lot of traffic there at bargain CPCs.
- You’re absolutely crazy if you aren’t bidding on traffic on parked domains. Try it yesterday!