Sep 282010

Negative Keywords Do NOT Affect Google AdWords Quality Scores

Lately, I've been reading too many pay-per-click articles that preach adding negative keywords as a means of boosting your AdWords Quality Score. We've been told for years now that this is absolutely false!

- Negative keywords do not affect your Google AdWords Quality Score.

Just to be sure, we asked Google if they would provide a statement, and received one from a Senior PM on Ads Quality:

"It's still true that negatives won't impact your Quality Score at any level. They are a great tool for boosting ROI but are not a useful means to improving Quality Score for a keyword, creative or account."

On a similar note:

- A keyword's matchtype will not affect Quality Score.
(I've read posts that claim this as well)

Here's the most helpful Quality Score document I've ever read (written by Frederick Vallaeys of Google). Make note of Slide 5, where they explicitly note how Quality Score is influenced at a keyword level.

While we're on the subject of Quality Score, let's call out one other important consideration:

- Quality Score is normalized for position.

There are many claims that you can bid your way to a higher Quality Score. The hypothesis is that your CTR naturally increases as you move higher on the page, and by increasing your CTR, you will obtain a higher Quality Score. Wrong! While it's true that you will likely experience a higher CTR by virtue of being higher on the page, it is not likely that you will increase your Quality Score by simply increasing your CPC bid. Google has enough data to expect a certain CTR level for qualified ads based on the query, keyword and position of the ad. (See slide 3 of the document I cited above for more details)

One of my favorite quotes from our Google rep is, "When you hear people talking about Quality Score and Tricks in the same sentence, it's bad information by definition". There simply is no way to "game" the system. Negative keywords and refined matchtypes are very important for a mature program, and I don't want it to seem like I'm arguing against that. Both refinements will lead to a greater ROI, so yes, please continue to use them! However, don't expect either to positively or negatively affect your AdWords Quality Scores.

Update 12/10/10: Recently I noticed this brand new page in the AdWords Help section:


46 Responses to "Negative Keywords Do NOT Affect Google AdWords Quality Scores"
Linda Bustos says:
Great to hear from the horse's mouth, but don't negative keywords help your QS indirectly, by improving your click through rate? (If you can weed out as much irrelevant impressions as possible)
Hi Linda, That's precisely the purpose of the post. Negative keywords do not directly or indirectly impact Quality Score. The reason for this is that Google determines Quality Scores off of the exact matching impressions and clicks to the user query. Therefore, eliminating impressions for queries that do not match the keyword exactly won't translate to an improved QS. Related - advertisers shouldn't restrict matchtypes from say, Broad to Exact in an effort to improve QS. Surely, from an ROI standpoint, this can be beneficial, but here we're stressing the fact that it isn't a path to improved QS. One thing I'd like to see from Google on this is an additional column reported in the interface. Every keyword would have an overall CTR (as it does currently) but would also have an exact query CTR metric as well. You can sort of get to this stat in a roundabout way via the Search Query Report, but I'd like to see it more predominantly displayed. Thanks for your question.
It bears repeating, Negatives are essential for paid search performance, but it doesn't have anything to do with Quality Score impact.
Thanks for publicizing these often-mistaken facts. Question I've been meaning to ask someone at Google - in case you're still in the conversation: Shouldn't the negatives impact (improve) the lifetime CTR for the account and the display URLs (slightly) in a way that does impact (slightly) quality score? It would seem impossible for them to filter out this effect. They'd have to guess the CTR of the impressions that didn't happen - that seems hard even for Google:-) In the weeds I know, but it would seem like in these slight ways negatives would have an impact. Would love any clarification you could get.
Hi Craig, Thanks for the additional thought/question. I too had that same thought ... "While they may only use the exact matching keyword and query combination, don't those other impressions have an impact on at least the account-level QS?" So, I asked explicitly about it (as you did) and their statement claims negatives "won't impact your QS at any level" - (account, creative or keyword). As for the calculation being hard, even for Google. I'm sure they have those stats by keyword on file (as they exist within the search query report) and can just summarize the granular data for the creative and account levels. Perhaps through this post / discussion, Google will offer more visibility to their coveted QS calculations :)
Craig, Matt, If they're only calculating based on exact matches, they just roll up THAT data to the AdGroup, Campaign and Account levels. It does seem that having a fully built out tail might help the more generalized QS calculations just because a larger volume of search would be matched to perfectly targeted KW. But I don't think excluding broadly matched data would make the calculations difficult. Am I missing something? George
Jeremy Brown says:
Google gets paid when a user clicks. Hence it makes sense to optimize your CTR. Google uses a number of factors (some disclosed, some not disclosed) to determine various Quality Score levels. I would be surprised if account CTR was completely ignored. How often do you think people at Google say, "here's all this data we have...let's ignore it!"
Although Google control for the effect of position, it's not a perfect system. They know how ads for a particular keyword might behave in different positions, and expect that level of change as you move up or down the page. But they don't necessarily know how "your" ad will perform in each position. Scenario: you have a great ad. Really great compared to the competition. If you move down the page, you would expect your CTR to fall by less than the other ads on that keyword would. So your CTR would be relatively higher in a low position than Google expect. In this case, moving down into a lower position but maintaining your high CTR would result in a better quality score. The inverse is also true: if you have an ad that is unusually affected by your position above and beyond the other ads on that keyword, you can expect your quality score to rise if you move up the page, as you will find a bigger effect than Google control for. The trick is to know how your ads respond, and factor that into your other decisions on your bid (conversion rate, conversion value, margin, etc).
David says:
So long as you consider what the Google rep says is the complete truth. There is no reason for them to say yes to any guesses related to QS and every reason to stick to the officially decided line. Best is to test it out and then decide. For me, it seems to have a positive effect.
There is one thing that I don't understand about how quality score is calculated. I think this is best illustrated by an example: I have an ad group containing one broad match keyword, "blue widgets". I have also added "blue widgets" as a negative exact match keyword keyword for this ad group. So the keyword should never be shown for its exact match variant. How is quality score calculated in this case? I've emailed my Google rep about this, but I'd be interested to see what everyone else thinks about this.
Richard, You bring up a very interesting case, one that an analyst on this end was also curious about. Our suspicion is that this keyword would receive the account level quality score, or some broader score, since it wouldn't have any exact query match data to go on. I'd suspect most folks haven't encountered this scenario since keywords they are adding are generally those that they DO want to receive exact query impressions for. Also, I don't know of any specific scenarios where this would largely benefit the advertiser, but hypothetically, if you had a very strong account quality score, you could try this out and see if a very general phrase could receive broad match (or non-exact query match) traffic at a higher QS level. Then again, it would mean that you could not have that exact matching phrase running whatsoever. As for the other comments about whether or not Google uses the account level CTR at all, I don't know if we'll ever know. One point I'm bringing up by this post is that it IS in the advertiser and Google's best interest that they calculate QS on exact match impression and click data only. If they didn't, why would an advertiser risk broad matchtypes at all? Google would be penalizing advertisers for running broad matchtypes, since they would be the ones to get superfluous clicks and impressions on keyword where Google believes there is a close match. In general, this will not increase overall CTR. Since that "expanded match" logic is ever evolving on Google's end, they don't want to slash advertiser's CTR as their systems determine what is and isn't relevant. The only area they can be certain of for QS is on the exact keyword that the advertiser supplies in their account. This holds the advertiser accountable for the keywords they are choosing to buy, but not those additional "expanded match" queries that Google is assessing.
I think the normalization provided by the ad-level QS allows them to calculate Ad Group, campaign and account QS by simple weighted averages. CTR is NOT the right metric to use in aggregate, for the simple reason that what is a "good" CTR depends on the keyword. Higher CTRs are expected on more targeted keywords, so a 3% CTR for position 3 might translate to a 9 QS for one KW, where another would need a 5% CTR to score that well. That's all baked in to the ad level QS, so no need to look at CTR at the more aggregated levels.
Matthew, Thanks for the response. For some head keywords I run broad/phrase/exact in separate ad groups with negatives set up so that I know which sort of query is going through each ad group. This may not be the best way of doing things, but it suits my way of working. For one of the accounts I manage I have the following QS for different match types of the same keyword: Exact: 7 Phrase: 10 Broad: 10 I'm not sure if this confirms your hypothesis or if it doesn't really say anything
I agree in general with what you are saying about the non-relationship between negative keywords and Quality Score, but I think there is a way that negatives can have an indirect, or supporting, role in boosting QS. I'm thinking of the peel and stick technique where you put the exact match of a broad match keyword into it's own ad group with an ad highly relevant to that keyword, then put the exact negative into the original group with the broad keyword. This should drive all the exact match traffic for that keyword to the new ad group, which since it has a more focused ad, will eventually get a higher QS. I've fleshed out this hypothesis more at
Hi Mark, It's an interesting idea that you present, and one that is likely to help control the appropriate ad that you'd like AdWords to serve for a given query. While negatives are a good mechanism for that (and I'll reiterate it again - negatives should not be ignored) they still will not have any bearing on QS. Prior to any 'peel and stick' technique being implemented, the keyword's QS, regardless of matchtype or negatives, was calculated by queries matching that keyword exactly. What your technique seems to be eluding to is the QS impact related to the keyword & ad creative pairing. It is true that each keyword and ad creative may share a different QS. Therefore, even though all QS calculations are determined on exact matching queries, Google also determines if Creative version A or Creative version B was used in the auction, and each keyword and ad creative pair may have a different QS. In your example, I'd speculate that the introduction of a new ad creative version is what led to an increased QS, not any changes to account structure, keyword matchtypes or negatives. On a related note, Richard's example of the same keyword on 3 matchtypes might also be impacted by this same phenomenon. If I launch the same keyword on all 3 matchtypes with only one ad creative version, they should all share the same QS until the end of time. If I introduce a new ad creative version to one of the keyword and matchtype pairs, it's possible that it may begin to carry a slightly different QS. Google does not display QS stats for ad creative versions. However, if you have your campaigns set to 'Optimize: Show better performing ads more often', you can infer which ad creatives have a better QS by viewing which ones receives greater impressions.
I just want to ask, if their is any documents or papers that will prove that negative keywords wouldn't affect the adwords quality score? Thanks
I have asked Google about this: 1. For broad match keywords QS is determined on queries which phrase match the query. The example above is correct, but it does not make it clear that the performance for the query "blue electric guitar" will affect QS. 2. In the case where a broad match keyword is blocked from appearing for these sort of queries the QS is determined by "the historic performance of that keyword across all accounts and the general performance of your account". I think this is pretty much what Matthew described. I am fairly surprised by point 2. And slightly disappointed; my QS of 10 is no longer because I am awesome :-(
Jonathan says:
Great post, I feel like I'm playing Carmen Sandiego because I've been looking to an answering for AGES! I'm hot on the tail after reading this post... but one BIG distinction needs to be made... So here is the question. Let's say we have 2 keywords: blue widget (broad) [blue widget] (exact) Now someone searches for: buy blue widget Will this search query use the quality score based off: [blue widget] OR [buy blue widget] In other words, WHICH exact match is CTR quality score based on, the keywords exact match, or the search query exact match. Please think about this as it makes a huge difference to adwords strategy because if [blue widget] has a CTR of 5%, and [buy blue widget] has a CTR of 3%, then it would be more beneficial to NOT use 'buy blue widget' in broad, phrase or exact, but rather to just let it come up through 'blue widget' broad match because it will then have a better quality score. Looking forward to your thoughts, and thank you again for a great post.
Jonathan, the answer unfortunately isn't very straight-forward. A Keyword's QS is a bit misleading. QS is calculated for each user query, so the user query "buy blue widgets" will prompt a different QS calculation for each of those KW. Google calculates which of the two KW will have the highest value for QS * bid and throws the winner into the auction with your competitors. How Google estimates CTR and relevance for each KW for each user query is a trade secret, and the fact that they can do all these calculations for each query (heck, for each letter typed!) is mind-boggling.
Hi Jonathan, The QS is based off of exact match queries, so in your example, if the query is 'buy blue widget' and you do not have the keyword 'buy blue widget', this particular query will not impact the QS in any level of your account. As for which keyword variation Google will choose to serve within your account, you may find this help page to be useful: This page describes what happens when more than one keyword within your account matches (or is eligible to serve) for a given search query.
Jonathan says:
Hi George and Matthew, Thank you for your replies... however my post must have been unclear in some way as you've both missed the question. Let me try and add to my previous post. For a minute, let's forget all of the quality score factors, and just focus on the main topic of Matthews post, and that is the fact that when CTR is factored in for quality score, they use the exact match. Now my question is, when a broad match or phrase match query is made, do they use the exact match of the keyword, being [blue widget] in my example above. Or do they look at the CTR of other times [buy blue widget] has shown for your account. Do they look at the CTR of the keyword, or the search query? I hope that helps clarify it. If we don't know the answer to this question, then knowing that CTR is based on the exact match means nothing because we don't know whether it's the exact match of the keyword or the query.
If I understand the question, the answer is: for each unique user query, Google makes an estimate of what the click through rate is likely to be for each KW that is relevant to the query. So the QS for "blue widget" varies for each different user query. The QS listed as the QS for that KW is the exact match QS. That's what I meant when I said a KW's QS is a bit misleading. Does that get at what you're after?
Jonathan says:
George I think I see where you're going with it. Basically any broad or phrase match queries are based on the CTR history for that SAME query, not the keyword that triggered your ad to show. So if 'buy blue widget' was the query, and "blue widget" was the keyword, then the search for 'buy blue widget' would use the CTR history of 'buy blue widget' NOT 'blue widget'. So bascially, a search will ALWAYS use the CTR history (as a factor in QS) of that exact search query, not the keyword.
Really they're trying to guess what the CTR will be for the combination of this user query and each potentially relevant ad in the account. So comparing the CTR for serving "blue widget" against the user search on "buy blue widget" vs serving "buy blue widget" against that user search. They do a great deal of testing to make these estimates and of course it's not just the KW but the combination of KW and copy for each ad that they're looking at. Crazy complicated.
Hi Matthew, Do you agree with the quote from Clickequations: "When you add a keyword to your account and use a broad or phrase match type, you attract queries that are related or similar to your keyword, but quality score is not calculated for these queries. When the CTR of those queries is lower than that of the identical query, you get an undeserved boost. When the CTR of those queries is higher than that of the identical query, you pay a quality score price."? /Frederik Trovatten
George: The point you make in #22 above is one I've never heard before. If I understand what you're saying the suggestion is that they estimate CTR for every unique query, but they display QS for a given keyword based on the times that query=keyword. This is a huge distinction from the assumption (as I understood it) that the QS calculated for those identical query-to-keywords were applied to those non-matching queries. It's great to finally understand and have an articulation of the alternative. Now we just need Google to confirm which is true - night and day difference between them. I have some conversations with the right folks later this week, and will try to get someone to confirm and go on-the-record.
Craig, that is my understanding from conversations with folks who aught to know. If we think of QS as largely a proxy for anticipated relative CTR, and we recognize that Google is primarily interested in maximizing revenue per page view, it makes sense that they'd take pains to anticipate correctly given the context and the unique user query presented each time. Mind boggling that they can actually do all these calculations at scale and speed, but I'm pretty confident that's what they do. Let us know if you hear differently!
Thanks George. Your premise and statement of their goals makes sense, and if the execution is as you say not only is the technology impressive but the fine line they've walked in describing it all to the world thus far. They clearly deserve as many plaudits for PR and marcom as technology. I do believe (and hope, not sure what order) that I'll be able to pin them down on this very soon. Thanks for all your knowledge sharing. Happy Holidays.
Mike Seddon says:
Hi Matthew, Excellent article thanks. I heard an interview with Frederick Vallaeys (why do I always want to call him Franky Valley!) a while ago on PPC Rockstars. Your article and the slides nicely tie all together for me. Regards Mike
dan says:
Is this blog post still accurate, seems like Google is telling me different, my head hurts trying to work out which one is right. "However, not using negative keywords can mean that your ads are shown to users who aren't interested in your business or service. This untargeted traffic can lower your keywords' Quality Scores and hurt your return on investment"
Dan, good question! We'll circle back with our Google reps to double check, but I suspect we're both saying the same thing and the difference is semantics. When people talk about "Quality Score" they're usually talking about the number attached to the ad by Google. That number is an average score across all the searches for which the ad is served. It goes against everything we've heard from Google on the subject -- we've always heard the average only applies to traffic exactly matched -- but it's possible that this number would be lowered by poorly broad matched traffic, not filtered by negatives. The key point is this: average QS is irrelevant. Google determines the QS for an ad based on each individual query (user search, domain, time of day, geography, past browser behavior, etc), and it's that number calculated on the fly that determines your AdRank and actual CPC (as distinct from your bid). In other words, the fact that the ad has a lower or higher QS for other user searches is irrelevant for this particular search. We'll check to see if the help center answer is right or wrong with respect to the calculation of the average QS, but the bottom line is averages don't matter; it's all calculated for the individual user search.
dan says:
Thanks for the clarification and the great resource. I guess it makes sense to keep a tight rein on negatives in any case.
Exactly right, there are all kinds of good reasons to build out comprehensive but thoughtful negatives, but QS probably isn't one of them. George
Dan, our Google rep confirmed that RKG had it right and the help center post you referenced was wrong. The QS metric is calculated on exact matched traffic from only. They will fix the help center post. Thanks for pointing that out! George
dan says:
Thanks George for following up! Glad this is clear, otherwise it would seem a bit unfair that Google can match us up broadly with a ton of keywords that suck for us.


Check out what others are saying...
[...] Negative Keywords Do NOT Affect Google AdWords Quality Scores, Rimm Kaufman [...]
[...] This post was mentioned on Twitter by ClickEquations, ClickEquations, bgTheory, Ryan Beale, Rimm-Kaufman Group and others. Rimm-Kaufman Group said: rkgblog: Negative Keywords Do NOT Affect Google AdWords Quality Scores: Lately, I’ve been reading too many pay-per... [...]
[...] [...]
[...] Mierzejewski at the Rimm-Kaufman Group Blog has published an excellent post on why negative keywords do not affect Google AdWords Quality Scores. (He also debunks a few other cherished Quality Score myths.) In summary, he provides several [...]
[...] This post was mentioned on Twitter by Thomas Petty, Thomas W. Petty. Thomas W. Petty said: Negative Keywords Do NOT Affect Google AdWords Quality Scores [...]
[...] *Since this post, there has been a few changes and discoveries -  Matthew has also written a post and detailed how negative keywords impact Quality Score* [...]
[...] It is also important to aggressively research and add negative keywords. This will increase your CTR and reduce your exposure to those searching for something you do not provide. Consider running an AdWords Search Query Performance report daily or weekly, mining your server log files, or checking your Analytics account for negatives and add them to your campaign negative list. Matthew Mierzejewski has also written a fantastic post on this topic and detailed how negative keywords impact Quality Score. [...]
[...] It is also important to aggressively research and add negative keywords. This will increase your CTR and reduce your exposure to those searching for something you do not provide. Consider running an AdWords Search Query Performance report daily or weekly, mining your server log files, or checking your Analytics account for negatives and add them to your campaign negative list. Matthew Mierzejewski has also written a fantastic post on this topic and detailed how negative keywords impact Quality Score. [...]
[...] If you’ve read a little about quality score (QS), you know that the biggest factor that influences it is CTR. When we plan about increasing CTR, new ads in more focused ad groups come to mind, sometimes we even remember to include and keep expanding a good list of negative keywords (although we know that that has no direct impact on QS). [...]
[...] If you’ve read a little about quality score (QS), you know that the biggest factor that influences it is CTR. When we plan about increasing CTR, new ads in more focused ad groups come to mind, sometimes we even remember to include and keep expanding a good list of negative keywords (although we know that that has no direct impact on QS). [...]
[...] It is also important to aggressively research and add negative keywords. This will increase your CTR and reduce your exposure to those searching for something you do not provide. Consider running an AdWords Search Query Performance report daily or weekly, mining your server log files, or checking your Analytics account for negatives and add them to your campaign negative list. Matthew Mierzejewski has also written a fantastic post on this topic and detailed how negative keywords impact Quality Score. [...]

Leave A Comment