by
23 Comments

Google AdWords Using Non-Selected (and often Non-Targeted) Keywords?

posted on July 12, 2007


Today we were doing some experimentation at my firm, and we came across something very bizarre with Google AdWords. We were trying out Google’s new search query reports, which allow you to see the actual search queries searchers used to find your ad. This differs from the keyword reports in Google AdWords in that these reports show the actual queries versus the keywords that you establish in your ad groups.

Why is that helpful? From our perspective, we thought they might provide some interesting data and insight into additional negative keywords we might want to troubleshoot for. Indeed the report did just that. In fact, the report yielded many new negative keywords for us to consider that were not highlighted by Google’s negative keywords tool.

But what was really shocking about this new report is that we found that our clients’ ads have been showing for keywords that do not appear in either the campaigns or the entire account. What do I mean?

Let’s use a fictional example. If the client were Apple and advertises only on Apple-branded terms, this report showed that the Apple ad appeared and was clicked on for queries on the term Microsoft — even though Apple might not be advertising on that term. And what’s worse, in the examples we saw, our clients could be spending quite a bit of money on these terms.

Until this new report debuted, you couldn’t tell that (from our example) the Apple ad might be coming up for the term Microsoft because all of the previous reports provided by Google were focused only on the information that YOU gave Google, rather than Google providing the actual search query data.

Our theory is that perhaps Google is trying to “help” us spend more money by targeting additional keywords WITHOUT OUR KNOWLEDGE. Perhaps the Ad Bot found these terms on or near the landing page? Or perhaps Google is tracking data about “like searches” and serving up ads for queries it thinks are “alike”, such as an “Xbox” ad appearing on a “Nintendo Wii” search because searchers that look for the Xbox may also be interested in a Wii?

What also steams me is that we can’t seem to opt out of this capability (except through negative keywords perhaps). Why even establish keywords in an AdWords account if Google will just predict likeness of searches for you? I liken it to my Tivo. I love my Tivo. But most of the time, my Tivo doesn’t do a good job at making suggestions for me. A computer cannot replace a human brain or even, without a shadow of a doubt, predict its intent.

If you doubt what I’m saying here, just try a search query report in your own Google AdWords account. I’d love to hear what you find. We’ve tried it across several clients now and found some disturbing results.


Related Posts

Zemanta
  • http://www.mellonstock.net Steve

    I’ve noticed the same thing with some of my campaigns. The first thing I wondered was which keyword they were attributing the impression or click to. The expensive ones?

  • http://www.getelastic.com Jason Billingsley

    I have known about this for some time. I do live visitor tracking and often see keywords that are ‘close’ to my campaigns. Since when does “buy card” equal “shopping cart”?

  • http://www.wpromote.com Mike

    As of today, the search query report is very different. It no longer allows you to look at the keyword that was triggered by the odd search query. Also, if you ran any reports in the past, they are probably out of your reporting center by now. Google is well-aware of how imperfect their broad matching feature is and it seems they haven’t quite decided how to deal with the situation.

  • Pingback: The Search Insider » Blog Archive » Google Search Query Report Suddenly Changed

  • http://www.ppclab.com Evan

    Janet, This is nothing really new but people need to know about it. It is Google’s expanded match. I posted about it about a year ago with data when I first started seeing it. http://www.ppclab.com/2006/7/google-testing-expanded-adwords-relevancy .cfm

  • Jan

    Yup, I also noted this. I was running a campaign in Belgium in the dutch language. (Belgium has 2 official languages : dutch and french)

    I bid on beltonen which is ringtones in dutch. When I searched for sonneries (ringtones in french) my ad also appeared although this keyword was not in any campaign I was running.

    I asked google and they gave my the expanded match answer but expanding into another language is going a bit too far in my book.

  • http://www.apogee-web-consulting.com/ Richard Ball

    Most people (even search marketers) don’t seem to recognize that broad match has actually been *expanded* broad match for a long time. You have a few different options as to how you can deal with expanded broad matches. I detailed some of these a few months ago:

    http://www.apogee-web-consulting.com/blogger/2006/12/be-careful-with-a dwords-expanded.html

    One tip that helps when you’re building keyword lists – make sure “Use synonyms” is checked when using Google’s keyword tool. That’ll give you a good idea as to how expansive broad matches will be for a given term.

  • http://www.search-mojo.com Janet Driscoll Miller

    To Jan’s comment, (I did not get into this in the original post) we also had JAPANESE words in our search query results. I can tell you that a) we chose only English as our language for the campaigns and b) we are only advertising in the US. In Jan’s case, I MAY be able to see expanded match for a country that has two languages, but the US (while technically having no official language) should default to English (perhaps also Spanish), but certainly not Japanese.

    Also, if this is “expanded match”, why am I not able to opt out of this? I personally think this is a very underhanded approach by Google, and for the most part very hush hush. Seriously — how many general AdWords advertisers would be aware of this and that it is even an issue?

    Google needs to STOP THINKING FOR ME and try to do me any favors. If expanded match is the issue here, then it is obviously and seriously flawed.

  • http://www.acteva.com Greg Moore

    Google should rename broad match, to really, really, really, really, Reeeeealy broad match.

  • http://www.acteva.com Greg Moore

    Does anyone know, what’s the best free way to see all user search queries that trigger AdWords ads? The Google report doesn’t show all queries. I’d like to see everything.

  • http://www.warmsentiments.com Tim Daly

    Good analysis Janet!

    I first started seeing this back on June 22nd on the front end before the reports became available. This was Q&A I posted out at LinkedIn:

    http://www.linkedin.com/answers/marketing-sales/advertising-promotion/ internet-marketing/MAR_ADP_INM/58762-1978920

    What I uncovered is that Google is appearing to use referring URL information (i.e. your most recent search request) to represent similar ads even in instances where your subsequent search has no relation to your previous search.

    In my example I shared on LinkedIn, I did a search on “gift baskets” and then a search on “auctions”. WIthin the results of the auction search I received “wholesale gift baskets” ads presented. Obviously, completely irrelevant.

    Biggest concern I think we have to consider here is the impact on Quality Score. Given the unpredictability of the irrelevant ad serving, you cannot apply this to your negative list.

    BTW, since I uncovered this issue on June 22nd, I’ve seen my ROI performance drop precipitously. Anyone else seeing the same?

  • http://www.apogee-web-consulting.com/ Richard Ball

    Janet – ALL broad matches are potential expanded matches. You can NOT opt out of expanded matching. It’s been this way since 2003 but their expanded matching algorithm has become more expansive over time. During the past 12 months, it’s become more obvious that it is too expansive. Read this Google help page:

    http://adwords.google.com/support/bin/answer.py?answer=6324

    Notice that the explanation of broad match includes: “…as well as some related keywords and phrases via our expanded keyword matching technology.”

    Regarding how long expanded matching has been utilized, see this 2003 AdWords news archive:

    https://adwords.google.com/select/news/oct03.html#matching

    Again, ALL broad matches are actually implemented as expanded broad matches in the Google AdWords system. This is not new. What is new (over the past 12 months or so) is how expansive the expanded matching has become.

    One last point – in some cases, expanded matching is actually “contracted” matching. For example, you might have a multi-word keyword in an ad group list like:

    charlottesville hotels

    and Google’s expanded matching algorithm might display your ad on the single keyword:

    hotels

    Imagine what that could do to your ad’s performance? Spikes in clicks from untargeted traffic like this can be disastrous for small businesses. IMHO, broad match as expanded match (and contracted match) is broken. There should be a way to do broad match as originally designed.

    Caveat emptor.

  • http://www.getelastic.com Jason Billingsley

    Absolutely. Our average keyword price rose from $1.05 to $1.70 over the last few months. The quality score must be affected by the bad broad matches. Bounce rates are much higher on the ‘intelligent’ broad matches.

  • http://www.beyondthepaid.com Mel66

    Janet, regarding your comment about Japanese searches… This has happened to us too. I blogged about it here:

    http://beyondthepaid.blogspot.com/2007/06/google-adwords-search-query- report-good.html

    I asked my Google rep about the foreign characters, and this was the reply:

    “Regarding the foreign language queries, individuals are typing these words in foreign languages while his or her language preference for Google is set to English. Your ads are appearing for these queries because our system determines that these queries are relevant based upon expanded broad matching.”

    I think that’s, well, errant thinking on the part of Google, to be polite. But that’s their story and they’re sticking to it. Meanwhile, I have to keep adding these foreign-character keywords as negatives. Luckily there aren’t too many of them.

    My personal opinion is that Google (and the other engines) should have 4 match types: expanded broad, “classic” broad, phrase, and exact. That would give us the option of showing for tail terms, but not these crazy instances of expanded broad match going way too far.

    Melissa

  • http://www.prezzatech.com Brian

    yep i’ve seen this too. synonyms match up with broad matches. I try to avoid the broad matches if all all possible.

  • http://www.miketheinternetguy.com.blog Mike Belasco

    I have seen so many crappy broad matches I cannot even begin to explain. Finally when you think Google releases a tool to improve quality and improve ROI for its customers, they pull it out from under you. I call shenanigans on this one! (and the other negative keyword tool they seemed to sterilize this week)

  • http://www.apogee-web-consulting.com/ Richard Ball

    @Jason (#13) – FYI, expanded broad matches do not impact Quality Score. Google is clearly aware that their expanded matches are not necessarily relevant (which is rather ironic). Read this Google help page to see what I mean:

    http://adwords.google.com/support/bin/answer.py?answer=6137

    Relevant quote: “Also, your ads’ performance on keyword variations doesn’t influence your keywords’ Quality Scores, minimum cost-per-click (CPC) bids, and ad positions.”

    However, since CTR is one of the more important factors in Quality Score (both the QS for min bid and the QS for ad ranking), I’m not sure how their system separates impressions (denominator in CTR) from expanded broad matches (keyword variations) from “regular” impressions. IOW, the CTR they use for QS calculations must not be the same CTR reported via your AdWords interface. There must be an internal CTR that ignores expanded broad match impressions. Either that, or expanded matches do actually impact QS.

  • Pingback: Link dump: Articles for better website performance | SEO Internet Marketing Blog

  • http://ppchero.com Amber

    I think if you’re super concerned about getting irrelevant searches set your keywords to exact or phrase match.

  • http://www.search-mojo.com Janet Driscoll Miller

    Amber, I don’t think that’s an option for most people. Here’s what I mean…

    In one client’s case, we have a client that sells a brand of case called a “storm case”. It’s a shipping case that is rugged and watertight. However, in using storm case as a keyword on broad match, Google matches the word “hurricane” to “storm” as a synonym. Frankly, a storm is not always a hurricane, but that’s a discussion for another day. This matching serves our ad for searches on “glass hurricane candle holder”.

    Now yes, I could do only phrase or exact match. However, I then severely limit my ability to find new keyword combinations. For instance, what if someone searches for “storm shipping case” or “storm waterproof case”? Phrase and exact match would not allow for that — they can severely limit your visibility AND they often cost more per click.

  • Pingback: Search Marketing Sage » Blog Archive » Broad Matching Your Brand on AdWords? Think again!

  • Pingback: Search Marketing Sage » Blog Archive » Mining the “Long Tail” of Google Search Query Negative Keywords

  • Pingback: Search Marketing Sage » Blog Archive » 6 Completely Obvious Things I Know About Google Adwords that the Average User Doesn’t Seem to Know…

Filed Under: General