Analyzing Search Terms: But We Already Have an Adgroup for That!

This essay is going to deal with a specific problem many users of Google Adwords face. What happens when you are using the “See Search Terms” feature and you notice that the search queries are triggering the wrong Adgroup?

Here is a real world example straight out of our Bridal Party Tees Adwords account: for a while we had been noticing that our “Bridal T-Shirts” Adgroup was being triggered by search terms such as “Bachelorette Party T-Shirts”. This was particularly frustrating because we already had a different Adgroup focused specifically on bachelorette party t-shirts. This adgroup included all keyword variations of “bachelorette party t-shirts”, it had ads specifically for that product, and most importantly, it had a landing page specific to that term. People were searching for bachelorette shirts and even though we had a beautiful landing page built just for them, they were being taken to our landing page for bridal t-shirts.

In fact, this was happening all over our account. Almost every adgroup was being triggered by keywords from adgroups we already had in place.

Despite this, our margins were still excellent. I could see that people searching “bachelorette party t-shirts” were still converting at a good rate, even if they clicked on one of our Bridal T-Shirt ads. Perhaps the rate was not as good as if they had been shown the correct ad, but still pretty darn good. I assumed this was because the site itself is relatively simple. We sell custom t-shirts to bridal parties and bachelorette parties and that’s about it. (Well, we actually sell a lot more than that, but those are definitely the majority of the business.) I also assumed the good conversion rate was partly due to our easy-to-navigate menu (designed by me, by the way /humblebrag), which lets users easily jump from bridal to bachelorette.

Because of this, we chose to let it ride. The only thing I could think to do would be to add negative keywords at the adgroup level, thus forcing the correct ad to show up for the correct term. But the Adwords auction machine works in mysterious ways, and I was afraid adding those negative keywords would result in a loss of volume.

This went on for a while until one of my newer team members was going through the Bridal Party Tees account, adgroup by adgroup, and using the See Search Terms feature. This particular team member was rather anal about keywords, which is the exact quality you want in a paid search specialist. The fact that search queries were triggering the wrong ads was driving her nuts. I couldn’t blame her, it had always bothered me as well.

We finally agreed to take the issue to our Google Adwords Reps. They advised us to use negative phrase match keywords in every adgroup. For example, add “bachelorette” as a negative keyword in our Bridal T-Shirts Adgroup. Then they told us to wait a couple days while they double checked with their colleagues. A few days later, they reconfirmed: add the negative phrase match keywords. So we did.

Over the next week, our account tanked. Impressions were down, clicks were down, and revenue was down. The difference in our margins was not significant. We had made a mistake.

We immediately deleted all of the negative keywords we had just added. Eventually, things went back to normal.

So what happened? First, let me tell you how our adgroups are structured. We are very precise with our adgroups. Each one represents a single, specific concept such as “bachelorette party t-shirts”. We have different adgroups for bachelorette apparel, bachelorette tank tops, and custom bachelorette t-shirts. Within each adgroup, we have every keyword duplicated three times, once for each match type. In essence, each adgroup is three individual adgroups, consisting of broad, phrase, and exact match types. We adjust bids by match type and create very specific ads for those keywords.

For the most of our keywords, our quality score is a 7 out of 10. But throughout our Bridal Party Tees Account, we actually have a good amount of 9s and 10s too. And here is where we went wrong with our negative keyword plan: we underestimated the importance of these quality scores.

I said earlier that the Adwords Auction machine is a mysterious beast, and it is. No one knows the details of how it works, but this is the general idea: let’s say I have a quality score of 10 and my competitor’s score is 5. If we both bid $2.00, then my total score is 20 (10 x 2 = 20) and his score is 10 (5 x 2 = 10), which puts me in first place. If he wants to beat me, he has to more than double ($4.01!) my bid to win the auction and get to first place. (Again, this is roughly how it works.)

It is my estimation that our broad match keywords with a quality score of 10, sprinkled throughout our Adgroups, were winning a lot of auctions for keywords from a different Adgroup. They were beating not only our own exact match bids for that keyword, but our competitors’ bids as well. Once we put the kibosh on that, our competitors started winning more auctions. (I use the word ‘winning’ as short hand for ‘ranking well’.)

Lessons learned:

  • If your ads are being triggered by keywords from a different adgroup, and those search queries are still profitable for you, don’t sweat it.
  • Your Google Reps will almost always have good advice when it comes increasing volume. But not always.
  • Pay close attention to your quality scores, especially if you are about to force the right adgroup to show up for the right search query.

Banner Advertising: Why Fansites Work and Facebook Fails

This is my first post on banner ads, and as such it is also going to serve as a bit of an introduction into how we use banner ads at eRetailing.  (At eRetailing, customers use our design centers to add text and art to apparel and other products, which we then print on demand. To learn more, ready my post on our flagship site, Customized Girl.)  If you want to skip to the Facebook vs Fansite argument, scroll down a bit.


I do a lot of banner advertising on the Google Display Network. I’ve used all kinds of networks over the years, including AdBrite, Chitiqa, and others, but Google has always been the best.  This is because Google has the largest inventory (thanks to Adsense) and the best method of targeting: keyword (aka contextual).

At eRetailing we make sure every ad dollar we spend results in a direct Return On Investment.  Maybe someday we’ll engage in more “brand” advertising, and we’ll use TV or billboards or Yahoo to try to sink into the public subconsciousness, but right now it’s all about making the most money possible with each individual dollar. Luckily for us, the internet exists which makes this super easy.

[Briefly: we’ve tried other forms of targeting. Network targeting (bidding on specific domains) can be just as good as keyword when used correctly. Retargeting is ok.  Not surprisingly, it works better the more specific you get, such as targeting visitors who added something to the shopping cart but never checked out. But beware, retargeting is a terrible idea if you’re paying per conversion and the ad network is counting “view-through” conversions.  In this case, you are essentially paying a third party simply because your company has repeat customers. Behavioral targeting, like the kind Yahoo uses for banners, is simply for not us.  Yahoo wants you to spend a huge sum of money up front so that you can “back in” to your targeting goals. Even once the “backing in” occurs, the ROI probably won’t match up to keyword targeting.]

The key, for us anyway, is using image ads to plant a very clear idea in our viewers’ minds. We prepare our users so that they have a very specific set of expectations and when they click on one of our ads we immediately satisfy these expectations on the landing page.  We do this by featuring the design on the item (often a t-shirt) as largely as possible on the banner. The design (like “Matt’s Tight End” on a pair of hot shorts or “Welcome Home Soldier” on an army wife t-shirt) needs to be simple with big block letters and high contrast.  It needs to be “immediately readable”, which is to say, if someone glances at it out of the corner of their eye, they should immediately understand what it is.

[Briefly, again: in my experience, this is not achievable with text ads on Google’s Display Network. We’ve run many tests with display text ads and have earned a (nearly) zero percent conversion rate. Text ads simply can’t build the perfect expectations in our users’ minds the way an image ad can. I also suspect most clicks on text ads from the display network are accidental.]

Our best results using keyword targeted banner ads, without question, have come from fansites.  By fansites, I mean any site that was built to serve the fans of one particular concept, from fans of running, to fans of the latest video game, to fans of young adult novels about vampires.

[Briefly, for the last time: if we ever decide to advertise on a particular piece of pop culture, we go out of our way to avoid copyright infringement.  We only cater to fan-created ideas, like Team Edward or Team Jacob.  No one owns the copy right to a first name like Edward, right? Right. Unfortunately someone did eventually purchase a “Team Edward” trademark, and at that time we immediately stopped all advertising related to vampires.  I think it’s pretty silly to stop your fans from showing their support for you, but that’s a post for another day.]

It’s great to advertise on fansites because fans go there to completely geek out about a certain topic.  While they are there, they are devouring everything they can about whatever interests them.  If your ad is relevant, it essentially becomes part of the site content. Fans are willing to look at your ad and think “Hey, that’s something I want!”

Interestingly, Facebook offers a similar targeting system. When you create your ad on Facebook, you can say “Okay, I want this to show up for anyone who is interested in running, or the latest video game, or young adult novels about vampires.”  We can reach the same people who are devouring our ads on fansites! It’s great! Except it isn’t.

It’s easy to think that the same surfers would be happy to see your ad wherever they are on the web, but that’s not the case.  People are on Facebook (and formerly MySpace – yes, we tried that too) to communicate with their friends.  They’re planning trips. They’re wishing you a happy birthday.  But they are not geeking out over one of their passions.  And they certainly aren’t shopping.

Here are some examples of banner ads and their corresponding landing pages:


Update: June 15, 2016. Well, in May of 2011, this post was pretty accurate. Of course, things have changed since then. Facebook has rolled out all kinds of great tools for advertisers, especially their remarketing pixel and their lookalike audiences. I do a lot more advertising on Facebook these days, but I don’t always enjoy it. There are some major issues with the way they report conversion metrics. Specifically, I think Facebook knows exactly who is about to checkout on your site (because they visited a page with “cart” or “checkout” in the URL) and then they display an ad to that user, and claim “credit” for the view-through conversion. But alas, that’s a post for another day…