Why Is Google Penalizing My Exact Match Keywords?

SMS Text

Once you’ve spent well over a decade mastering the art of the ideal pay-per-click campaign structure, you tend to start taking certain so-called “best practices” on faith. Concepts like device-segmenting and geo-segmenting becoming no-brainers, not even worth the time it takes to question whether or not they are a good idea. However, a recent issue with a new client has made me rethink and question everything I thought I knew about PPC.

This problem started after I sold a new client on my services, which revolved around creating the ideal campaign structure.

“It is important to hyper-segment your campaigns so that you get the most out of your advertising spend,” I said proudly. “Customers in New Orleans will respond differently to your product than customers in Des Moines, and therefore, your campaigns should be segmented so you can bid differently in one region than you would in another.”

I really sold it. The client was impressed. I then went on to talk about the trivial nature of keyword match types.

“Exact match keywords are exactly what the user typed. If the keyword is extremely relevant to the ad it serves and the landing page is extremely relevant, also, then Google will reward you by raising your quality score for that keyword, which will lower your cost.”

I was on a roll…

“Exact match keywords are going to be much more relevant to your ad and landing page than broad match, because you’re targeting exactly what that user is looking for with your keyword, ad, and landing page.”

Damn, I am good…

So the client was sold, and I began the process of building out their campaign based on what we had discussed. That is where we ran into some issues that I’m still trying to sort out.

I created, as I always do, segmented campaigns. Each campaign targeted only the U.S. I like to start there and segment by state, then city once I get enough data to start selecting the highest performing areas for the particular product or service. I created separate campaigns based on device type: a computer campaign, a mobile campaign, and a tablet campaign since people behave differently on each device.

I then segmented by keyword match type. Each campaign was exactly the same with the same ad groups and ads. The only difference was device targeting and keyword match type. The types I selected were modified broad match, phrase, and exact. I knew there would be some natural crossover, but I just wanted to get the campaigns live, and I would sort out any duplicate targeting issues later.

Here is a breakdown of the campaigns I created:

  • U.S. Non-Brand – Computers (Modified Broad)
  • U.S. Non-Brand – Mobile (Modified Broad)
  • U.S. Non-Brand – Tablet (Modified Broad)
  • U.S. Non-Brand – Computers (Phrase)
  • U.S. Non-Brand – Mobile (Phrase)
  • U.S. Non-Brand – Tablet (Phrase)
  • U.S. Non-Brand – Computers (Exact)
  • U.S. Non-Brand – Mobile (Exact)
  • U.S. Non-Brand – Tablet (Exact)

I thought about adding a negative exact match keyword to the phrase and broad matches but decided against it since I just wanted to get some traffic going first before refining the campaign. What happened over the next several weeks startled me.

I will say again that every campaign was an exact duplicate of every other campaign. The only difference was the device targeting and keyword match type. The first thing I noticed was that my cost per click on the exact match campaigns was almost double the duplicate modified broad match campaigns. This really bothered me because exact match is supposed to perform better since you’re targeting the exact search term. At least, that’s what I thought.

I dove in further and discovered that Google was apparently penalizing my campaign structure. Specifically, it was penalizing my exact match campaigns and rewarding my modified broad match campaigns.

Here is a breakdown of what my average quality score looked like:


Average Quality Score

U.S. Non-Brand – Computers (Modified Broad)  


U.S. Non-Brand – Mobile (Modified Broad)          


U.S. Non-Brand – Tablet (Modified Broad)          


U.S. Non-Brand – Computers (Phrase)                  


U.S. Non-Brand – Mobile (Phrase)                        


U.S. Non-Brand – Tablet (Phrase)                          


U.S. Non-Brand – Computers (Exact)                    


U.S. Non-Brand – Mobile (Exact)                          


U.S. Non-Brand – Tablet (Exact)                            


Obviously, Google favored my modified broad match over my phrase match and favored my phrase match over my exact match.  This is exactly the opposite of everything I ever thought I knew about PPC marketing. This prompted me to satisfy my curiosity by testing out a duplicate campaign that targeted just broad match (without the modifier).  The results after just a week were stunning:


Average Quality Score

U.S. Non-Brand – Computers (Non-Modified Broad)


U.S. Non-Brand – Mobile (Non-Modified Broad)      


U.S. Non-Brand – Tablet (Non-Modified Broad)       


Quality score is directly tied to what you end up paying per click, so it’s no surprise that my average cost per click for the new non-modified broad match campaign dropped considerably. In some cases, I was now paying only $1 per click, whereas I had been paying upwards of $3 for the same modified keyword. Another interesting note is that my average ad position jumped up several positions. I now had the #1 ad, instead of the #2 ad or the #3 ad, for most keywords, and I was paying a fraction of what I had been paying for more specific, modified, broad match keywords.

So why would Google penalize you for targeting exact match, phrase match, and even modified broad match, while favoring non-modified broad match instead?

Perhaps their recent default “feature” that targets synonyms and misspellings on phrase and exact match keywords is a hint. It’s apparent that Google doesn’t want you to hyper-target.  They don’t seem to care that exact match, targeted, long-tail keywords will obviously perform better for most businesses that are focused on leads or sales than broad match, one-word or two-word keywords. The fact that you typically can’t target a keyword made up of four or five words because they will immediately deactivate it based on “low search volume” is a perfect example of that. If I have a “George Foreman Super Champ Power Press Grill” for sale, then why should I be prevented from targeting that exact term?

It appears instead that Google just wants you to build some poorly structured, broad match campaigns, trust their better judgement, hand over your credit card, and shut the hell up. That’s the impression I got from their customer support, which kept reciting the same tired script over and over to me whenever I called.

Unfortunately, this article ends on less of a solution than I’d like to offer. Basically, you have two choices: advertise on Google and do whatever they want you to do or skip search engine pay-per-click advertising entirely. Bing’s market share is still only a meager 15% in the U.S. with only a 4.4% global market share, and I don’t see any other competitors, other than Baidu, who are useful if you’re targeting China.

So we are stuck with Google favoring their profits over the success of our campaigns. The solution? Stop segmenting traffic. Build a non-segmented campaign. Group all the device types, all the keyword match types, all the regions you sell in, and don’t even think about day-time parting! Let that campaign build traffic and quality score, and then, maybe, just maybe, you can “test” segmenting it out after maybe six, nine, or even twelve months.  Perhaps by then Google will have figured out that it can’t screw its customers for long before they cancel their credit cards.

Get the latest news from Search Engine Journal!
We value your privacy! See our policy here.
  • http://www.redengine.ie Michael

    This is really strange and is basically the complete opposite to what Google have been saying in recent months, getting PPC users to specifically target traffic through better optimization of accounts

    • http://mikeboudet.com Mike Boudet

      That’s why I wrote the article. o_O

  • http://www.motivationals.org Uncle Demotivator

    BUFU should be new Google motto.

  • James

    Do you have any data on conversion rates for the different segments and how they would affect your decision making in tweeking your campaigns (despite Google’s QS)?

    • http://blog.mikeboudet.com Mike Boudet

      interesting question. The quality of exact match traffic appears to be way better. I don’t have any data right now to support that, but I have been running the broad match campaigns for a while and it seems that I’m paying more for the same number of conversions. This works towards my theory that Google wants to sell more inferior-quality traffic, rather than highly-focused, excellent quality traffic. Why wouldn’t they? They can make more money that way and since they don’t really have any real competition, then who’s going to challenge them?

      • http://www.webholism.com Sara Thornton

        Glad to see that for good conversion rates the intuitive route works, which would make sense as consumers generally respond better to highly targeted adverts.

        As quality score is made up in large part from click through rate, did you compared the CTR between the two cases? If your CTR is lower for your exact match then this could explain the results, perhaps…?

        A final thought: as you were duplicating campaigns, and (from what I can see in your article) hadn’t yet set up the corresponding keywords as negative keywords for each AdGroup (e.g. for your broad match AdGroup, setting up the exact and phrase match of the same keyword as negative keywords, so that you didn’t inadvertently double target the same keyword), could this potentially have also contributed to the odd results?

  • http://www.jessekanclerz.com/blog/ Jesse Kanclerz

    Interesting data you’ve presented Mike.

    I decided to check the average quality scores in my mirror campaigns and found that the average is the same for the broad and phrase match campaigns but the exact match QS is a difference of 2.3 higher.

    Here’s a couple thoughts on possibly why I’m seeing the opposite results:

    1) The broad and phrase match campaigns I’m running have the negative exact match of the keyword.
    2) A longer standing account history. The accounts been running close to a decade but recently switched to using mirror campaigns.

    I’d be curious to see if the QS pattern you experienced with a new account holds if you add in the negative exact matches to the broad and phrase campaigns at the start.

    • http://blog.mikeboudet.com Mike Boudet

      I think you may be on to something with the account history…

      • Rob T

        Account history is absolutely part of the issue here. You’re account structure here is perfect, but it is the execution that needs some work. I assume you launched all of the campaigns at the same time and that is part of the problem. Higher level match types (broad, broad-modified) are always going to match to more queries and Google is going to favor these match types if all campaigns have the same amount of history (none).

        The solution is to launch all of the exact match campaigns first and let them gather history for a couple weeks. Once they’ve gathered some history and been partly optimized (ads, bids, etc) you should be able to launch the other match types safely. If the exact match campaign is performing well (relevant ad copy, good CTR), the higher level match types shouldn’t trump the exact match campaigns.

        Using exact match negatives in the broad and phase campaigns is also a good option that can help give the exact match campaigns a ‘head start’. I go back and forth on this, but it seems to be a good option if for whatever reason the phase or broad campaign gain traction more quickly.

        One other thing to think about is that maybe you’re phrase and broad campaigns are really just matching to queries that are more relevant to your campaign. If the keyword isn’t in exact match, you’ll see the higher level campaigns perform better. A SQR should help clear that up.

  • http://Search-mojo.com Tad Miller

    Just curious. What are your settings for Exact and Phrase match keywords? Are you including plurals, misspellings and other close variants or have you selected Do not include close variants. I’ve been really wondering about that new setting having the potential to kill exact match keyword quality scores.

    • http://mikeboudet.com Mike Boudet

      I have plurals, misspellings and other close variants turned off

  • Steve Werner

    This is strange; however, the fact is we’ve all been segmenting our campaigns, and continue to do so, because it’s effective. My exact match campaigns perform much better than my broad matches, with better QS and lower CPC. I can’t help but think there is more going on here than just Google shafting you.

    • http://blog.mikeboudet.com Mike Boudet

      How old are your accounts. have you tested this on brand new accounts?

  • http://www.cpasitesolutions.com Norm

    First I have a question. How do you get quality score reports specific to the decimal point. I though I only could get it rounded to the integer.

    Now I do have some thoughts to share. I don’t think you can account for all that goes into QS by comparing a 4.15 against a 4.56 and a 4.75 as shown for your U.S. Non-Brand – Mobile exact, phrase and broad keywords. Keep in mind it’s an average. So for the phrase and broad, some search phrases will give you a 2 and others will give you a 6.

    Google will now show you what the actual QS is for Search Phrases for phrase and broad match types, but they do show you the Cost Per Click in the Search Phrases report. This will give you a clue as to which broad and phrase match search phrases may be high quality scores, and you can then add those as exact matches.

    Also, just because you have an exact match, doesn’t mean that is the one google will display. For example, I have “accounting websites” as exact, phrase and broad match in one ad group, and I have “CPA websites” as exact, phrase and broad in another ad group. The ads in each group is specific to those keywords, just as Google wants and recommends.

    However, Google considers “accounting websites” as a broad match as “cpa websites”. The can and do show the broad match for that search phrase even though I have it setup as an exact match in a different ad group. I thought oh ya, sure, show the ad which will generate more revenue for you, but I found that was not the case. I saw examples where the cost per click was actually less on the broad match then it would have (apparently) been on the exact match. In my case, I added “accounting” and “accountant” as a negatives in my cpa ad group and I added “cpa” as a negative in the accounting ad group. In my case, I REALLY do want the CPA ad shown for cpa search phrases and accounting shown for accounting or accountant search phrases.

    Your cost per click is not determined (solely) by your bid and QS. It’s determined in larger part by the ad below you. So even though you setup your campaign to have it all be identical, that doesn’t mean the same ad is being displayed underneath yours for each search phrase. Even if your QS was also identical, that doesn’t mean your cost per click will be too.

    I had another thought, but it escapes me. If I think of it I will post again.

    • http://www.cpasitesolutions.com Norm

      Urgh, I wish you can edit replies for the first 5 minutes. “Google will now show you what the actual QS is for Search Phrases for phrase and broad match types” should have said “Google WILL NOT SHOW….” I’m sure there are other typos… 🙂

      • http://blog.mikeboudet.com Mike Boudet

        The answer is: I put the numbers in an excel spreadsheet and averaged them out. I appreciate the rest of your post, but I’m afraid I do not understand your point. My point is that if you set up two identical ad groups, one targeting broad and one targeting exact, with the same keywords and the same ads, and the same bids, Google will favor your broad match and send you a whole lot more crappy quality traffic through that then they will for your exact match, phrase mat and even modified broad match ad groups. The point is that they want to sell you more clicks, that are of substantially less quality. That’s a far cry from where Google Adwords started out and a damn shame.

  • http://www.ericmobley.net Eric Mobley

    I am just curious if this is an isolated incident or if more PPC folks are noticing the same behavior?

  • http://www.prunderground.com Alex Miranda

    I am so glad you posted this. I thought I was going crazy. The same exact thing happened to me and I could not understand why I was being penalized. I even created a new landing page and changed how the keywords appeared on the page. All along I was thinking it was me. I have been doing PPC for over 10 years now with great success and then this.

    • http://mikeboudet.com Mike

      Thanks, Alex. I posted this because I thought I was going crazy. Needed either people to tell me I’m crazy or verify they’ve sen the same thing. Looks like we got a little of both here.

  • http://www.webmarketers.com.au WMC Crew

    I think this has been gradually getting worse and worse with google limiting your control over how you manage and segment your ppc. I noticed about year ago google started to not like my granularity, moving traffic to adgroups it felt were more relevant.. If I want to test the difference between the conversion of the word ‘washer’ and ‘washing’ then I should be able to! Worst of all it’s so inconsistent from campaign to campaign.

    Then they wanted to limit ad rotations to 30 days.. Now lucky us.. We get 90.

    And now this… You are definitely right in your thoughts of how it is heading.. I’m starting to segment less as granularity is no longer as beneficial as it once was.

    • http://blog.mikeboudet.com Mike Boudet

      It’s a real testament that Bing converts better these days despite how totally incompetent they are. It’s a sign that Google figured out they could make a lot more money by selling less-quality, crap traffic, and reworked their algo to reflect that. It should be a crime to destroy the tool that created such prosperity for so many small businesses.

  • http://www.ignivasolutions.com/ Igniva Solutions

    I have no doubts over your analysis, Google is just like that. And guess you will learn one day it implements this kind of mentality in organic results as well in a way or two just to keep SEO people at tenterhooks. 😉

    • http://mikeboudet.com Mike Boudet

      Some would argue that Panda is just that.

  • http://www.cheapconservatories.net Jamie

    Isn’t this because your click through rates are performing relatively worse to your click through rates on adverts showing on exact match (which is probably because they’re more competitive hence competitors will have put more effort into perfecting ad texts for them) ?

    • http://mikeboudet.com Mike Boudet

      If that’s true than Google places a greater emphasis on how crowded the market is for a specific keyword, rather than how relevant it is for the ad and landing page. It’s just as counter-intuitive and goes against what they say their best-practices should be.

  • http://www.evemilano.com EVE Milano

    Very interesting data, I was asking myself the consequences of their recent default “feature” that targets synonyms and misspellings. Probably your conclusions are right, I’ll test it asap. Thanks a lot

    • http://blog.mikeboudet.com Mike Boudet

      You’re welcome. The only defense we have against a monopoly is the right to organize and share knowledge. I want to know if others out there are experiencing the same issues I am and figure out if there’s a better way to approach the problem. I actually recorded a podcast about this, which I hope will be the first of many…


  • http://launcheteer.com Robert Duncan

    Very Interesting, but basically your QS is going down because you are in a more crowded market. Meaning your CTR is probably down. This is the most important factor in your score.

    It doesn’t matter about your landing page (even though they say so). I run ads all the time where Google hasn’t viewed or cached my landing page, but yet I had a good quality score.

    Google’s best practices are ads that bring the most clicks. Of course Google wants the ads to be relevant but it is not likely you are going to be paying Google for ads that are not relevant, thus no need to really monitor your landing page.

    • http://mikeboudet.com Mike Boudet

      If that’s the case then why would the exact match be a “more crowded market”. Aren’t competing, not only with all the other savvy internet marketers out there competing on exact match, but also a million others who bid broad for the same keyword?

  • http://www.weddinghorses.com/ Horses

    @Mike. Just want to add my 2 Cents worth as I am a total non conformist on everything to do with SEO and PPC.
    You will find your PPC keyword relevence is algorithmised by Google according to business listings in the area you are targeting. IE, postcode and business density. If you use broad reach keywords, you are normally likely to pay more per click. Ignore their CPC advice also.

    • http://mikeboudet.com Mike Boudet

      Please explain. I read your post three times and don’t really understand what you’re trying to say.

  • http://www.mountainriders.com Troy Tanga

    Aloha Mike,

    Its been too much fun watching Google take my campaigns from $2000 to $3000 per month up in cost about 50%. Spoke to my account rep now in India suggested I remove miss spelling and close variants in the settings.

    Did so and my costs Increased even more!! Don’t want to just set a budget on a campaign that has been working effectively for 2-3 years

    The average Quality Score of my Search Campaigns is in the 9-10 range. Average CTR 7-8%. I have added approximately 3700 negative keywords, but everyday I have to add 50-1 more trying just to keep up.

    I have some keywords that are modified with the + and some with “”. Going through Adwords Editor and trying to decide if I should remove the modified phrase matched.

    Trying to figure out the best way to approach. Any ideas