However, I have now got a single adgroup with a single exact-match keyword, with one very old ad and 3 new ads (made/approved in the last week).
The CPC for exactly the same keyword varies across the three ads considerably:
Ad 1 - Served 90% - £0.20 CPC - established high CTR
Ad 2 - Served 1.1% - £2.35 CPC - new this week
Ad 3 - Served 1.1% - £2.17 CPC - new this week
Ad 4 - Served 7.8% - £1.90 CPC - new this week
(ad 2,3,4 have the same if not higher CTR than ad 1)
Surely this is clear proof that historical ad CTR is crucial in ensuring a reasonable CPC for a keyword?
Chris
From what I understand:
The CTR used in the 'ad rank formula' is based on the KW - not the ad.
The reason this is important is that when changing an ad, you're not losing ad rank position, because it's KW based.
However, in your example it really looks like the above can remain true, but it's the actual paid CPC is based on the ad CTR not the KW CTR.
I've not seen an example that far apart before in paid CPC. Before, a few cents here and there could be explained by different variables at the point that particular ad was shown.
I have seen examples where the ad with the highest CTR paid a few cents more than the lower CTR ads, thus making the slight variations just part of the time/advertisers/network the ad happened to show more.
However, in your example it really looks like the above can remain true, but it's the actual paid CPC is based on the ad CTR not the KW CTR.
This is not true since Chris999 indicated that
ad 2,3,4 have the same if not higher CTR than ad 1)
Since ads 2,3, and 4 have the same or higher CTR comparing to 1, the actual CPCs for the new ads should be lower.
Chris999's observation is very interesting and is a good example to question about my understandings on the AdRank formula and the actual CPC calculation.
As we know the AdRank formula is very simple: Max. CPC x CTR. However, it's very complex to understand. There is no problem with Max. CPC since it's a single value, which is in our control. The problematic is in CTR. How is CTR based? Literally, there is an infinite number of CTR. Is it based on the recent 1000 impressions for old established kw or the default value for a new kw? Are there other factors, which are also taken into the calculation such as the history of each ad (new ad with long established kw, 1 month, over 1 year)? To illustrate my confusion:
Long established Ad: the effective CTR is based on a certain percentage of the recent CTR (1000 impressions) + some percentages on the entire history CTR of the kw + others
Newly created Ad: the effective CTR is based on a certain percentage of the default CTR + some percentages on the entire history CTR of the kw + others.
I also hope the CTR in the AdRank and actual CPC algo be explained one more.
Thx,
FromRocky
I'm very interested in this topic though and like to know more about the relationship between individual ads and keywords. I have the feeling there must be more to this than the "keyword determines all"
I read the "ad 2,3,4 have the same if not higher CTR than ad 1" backwards this morning (i.e. have lower - not enough coffee when I posted).
Which makes your statements correct.
Actually, it makes me wonder if Chrisk999 is looking at a date range that's longer than the new ads ran, where it's possible that before the new ads were around, he ran ad 1 with a low KW Max CPC for a while and had a lot of clicks, thus dragging down the all time average for that ad in terms of paid CPC.
If that's the case, looking at a more recent time frame when every ad was run might shed a different set of results.
From what I understand:The CTR used in the 'ad rank formula' is based on the KW - not the ad.
I'm very interested in this topic though and like to know more about the relationship between individual ads and keywords. I have the feeling there must be more to this than the "keyword determines all"
This is a good topic, and I hope that you get a few responses from AWA on this one.
I'll do my best to collect a few engineers and drill down on this subject - and report back. Just to set the right expectations, I most likely won't be able to accomplish this today - as it is pretty hard to get those engineering folks to slow down for long. ;)
Anyway, I'll see what I can learn, and then post again ASAP.
AWA
Actually, it makes me wonder if Chrisk999 is looking at a date range that's longer than the new ads ran
Thanks AWA - I'd be really interested to know what the engineers say. A similar issue was also touched on by jamie_h [webmasterworld.com] here.
Apologies for bumping this, but I am still wondering if AWA had any success in investigating this issue?
No apologies required chrisk999 - in fact, thanks for bumping this. Your timing is perfect - as I have a meeting scheduled this afternoon to discuss this exact thing. I've done some digging to date, should get my info confirmed in a couple of hours, and hope to post with some solid info later in the afternoon.
Really, I owe you all an apology. Between one 'priority' task and another, I've been really remiss in following through, researching, and posting more info on this subject - which is an important one. Sorry.
You've been very patient indeed, and I appreciate it.
More soon.
AWA
Apologies for bumping this, but I am still wondering if AWA had any success in investigating this issue?
Ok, I've been able to get some info on this subject - although precise details of the algo, as usual, are not on tap.
The short story is that, yes, the ads positioning algo has in fact shifted from considering only keyword CTR, and now folds-in information about the relevance of the ad itself, including the Display URL.
By way of background, back in June 04, an announcement was made that incremental Ads Quality changes would be taking place over time, and the change we're discussing here is a part of that ongoing effort. It's based on the principle that 'CTR' is highly related to the targeting between keyword and ad, as well as to the 'quality' of the ad itself. (Those who organize by targeted Ad Group, and run tests of multiple ads in order to squeeze out the very best CTR (or ROI) will know the essential truth of this principle.)
I hope this brings some clarity to what some of you have reported seeing in your accounts, and again, I apologize for having dropped the ball in not getting this info posted sooner.
AWA
The short story is that, yes, the ads positioning algo has in fact shifted from considering only keyword CTR, and now folds-in information about the relevance of the ad itself, including the Display URL.
Care to elaborate a little bit more about this?
Does the ad CTR now play a part in the ad positioning and price structure?
Does the ads wording play a part?
How does G determine ad relevance? There are many terms where an ad can be relevant, yet not contain the keywords in the ad (in some tech fields this is due to the phrase being longer than 25-35 characters, so many abbreviations are used)
Care to elaborate a little bit more about this?
eWhisper, you can always be relied upon to quickly ask the 'depthy' questions - and as a result, you really know AdWords inside and out. ;)
In this case, though, for competitive reasons, I'm not able to elaborate on the algo - or to provide specific details as to how relevance is determined.
The essential message, however, is that, yes, the ad CTR does play a part in the positioning and price structure. That is what I was trying to get at with the quote below - and sorry if I wasn't clear.
The short story is that, yes, the ads positioning algo has in fact shifted from considering only keyword CTR, and now folds-in information about the relevance of the ad itself, including the Display URL.
AWA
So AWA was unable to elaborate too much (G loves their secrecy, after all), but nothing is holding me back, so I will let you guys know what I was able to squeeze out of our contact at Google regarding the KW-Ad creative interactions that some of us seem to be grappling with. The confusing aspect is that this officially applies to the predicted CTR (used in determining KW status) but seems to effect the AdRank as well -- I'd love to hear some thoughts or insight on this point. Anhow:
1) This factoring of the ad text into the AdRank/pCTR algorithm has been growing steadily over time, and G people are constantly tweaking/improving things, so you may notice changes in even short periods (this may help to explain short-term bizarre behavior as well). This began, from what we were told, with the rollout of the new keyword states not too long ago (On hold, in trial, etc.) and has been becoming more of a factor over time.
2) The interaction is not a history or CTR of the ad text itself, but the actual interaction of the KW and the ad text (As AWA pointed out). Some of it is quite logical, such as rewarding having keywords repeated in the ad text title, description and even display URL. I imagine that they are basically looking at system-wide trends of what ads "work" for particular keywords, and using these principles to guide how to reward (or punish) good ads -- or rather, well chosen keywords for a particular ad text.
3) So (and correct me if something looks off) the predicted CTR of an individual KW is basically pCTR = Recent CTR (last 1000 impressions) + All-time CTR + System-wide CTR for that KW + Some Ad-Text Value. The real mystery here is, of course, the coefficients of the components (which themselves appear to be changing).
3) The take-away for me was that in the past, we could change ad texts with no direct penalty on the KW, but this is no longer the case.
The bottom line here is to have solidly grouped keywords with relevant ad texts. While we will never know all the intricacies of the AdRank algo, if you follow general smart guidelines of running AdWords campaigns, people should be in good shape.
I'd love to hear other thoughts on this!
mothne
3) The take-away for me was that in the past, we could change ad texts with no direct penalty on the KW, but this is no longer the case.
This is a very big change.
1. We use to be able to split test, change ad copy, split test, change copy, etc on regular intervals without worrying that the changing ad copy would affect CTR/placement - no longer the case.
2. This changes a very common piece of advice: 'Make a second ad for search partners, wait for it to be approved, then delete the first ad - it won't hurt your CTR/exposure at all'.
3. When doing limited exposure 'branding' campaigns to capture the audience for clients who were on tv/radio/etc, rarely was the keyword used in the title. If Google starts weighing ads based on where the KW appears in an ad, this is a big change to these types of campaigns.
4. If display URL matters even a little, then we're going to start seeing more: keyword.example.com or example.com/keyword ad texts as it'll increase placement regardless of how it looks.
5. I hope this doesn't lead to everyone using dynamic insertion - I can't stand search pages where every title is the exact same - it doesn't help the user - but it looks like it might help in placement now (unless only the backup text is used to measure relevancy). I've always suggested to people, if everyone is using dynamic insertion - use a static title - using static with some sort of dynamic insertion (i.e. Title: Buy {Keyword:Anything} at Widgets.com) might be the new rule to follow.
Just saw these display URLs:
Widget.Find-Widget.Your-Widgets.com
Widget.Your-Widgets.com/Widgets
Of course, the display URLs resolve to 404s, but Google hasn't cared if the display URL does this for a while with subdomains.
Is ad copy going the way of the SERPs in widget.com/widget/widget-shopping?
In other cases, where I was advertising specific products, my display url would go to "widgets.com/specificproduct". In these cases, I did not see big position drops, in fact I have seen gains recently.
So, I think display url may be a pretty big factor. It seems silly to me though to optimise urls in this way though for many things. For instance if you were advertising New York Times subscriptions, you'd want all urls to more or less go to NewYorkTimes.com. To have a display url of NewYorkTimes.com/nytimes for those that searched for "ny times" and NewYorkTimes.com/newyorktimes for those that searched for "New York Times" just seems redundent and odd. There are many cases where human understanding would preside way over anything that could be optimised to determine relavence, which is why the CTR seems a much better determinant of relavance...just my 2 cents...
Widgets
widgety widgets for
widget lovers.
widget.com/widgets
would it have some higher base ranking from the get go due to "relavance"?
Do we not wisely use many synomyms in search phrases? Are these being penalized? I don't include the 1/2 dozen synomyns for one of the main products that I sell anywhere in the ads. That would just be silly. I realize of course that finding the bolded words from the search phrase in the ad leads to better conversions, however with 100 characters that is not always possible to accomodate.
Older or less technical people use certain terms to describe my products, younger people use more precise terms. Does that mean the ad rank for those synomyns is potentially affected? I've kind of suspected that actually based on certain trends I've noticed with what goes quickly into "trial".
The presumption of finding the keyword in the ads will just lead to spammier ads now that the cat is out of the bag.
patient2all
Bizarrely, this time I am paying a very reasonable CPC (40p rather than the ridiculous £2.30 in post #1).
So either they've changed the algo slightly since post #1, or ad text relevance actually is a new factor to consider (CTR's of the two sets of creative are roughly similar).
Thank you for your responses AWA - very good to get confirmation of the new factors. Much appreciated.
One question for you though: are we supposed to just weather the high-initial-CPC storm and hope it drops as the ad becomes more established, or is there something I am doing wrong?
Before this post came up, I would have thought they would have got us on this rule if KW = "ads":
Google AdWords Ads
Ads, ads, ads are:
Fast, easy, and effective!Gimmicky repetition in "Ads, ads, ads."
I just recently learned it's okay to "make up" folders in your Display URL. I assume you have no "greenwidgets" or "green_widgets" as an actual location on your server, yet that redundancy is now considered desirable in a display URL?
Sounds like this rule:
"Your Display URL must accurately reflect the URL of your website"
can be bent too. So only the domain and TLD must be accurate?
I'm going to try some tests too with this newfound freedom to what I consider "spam" the ads (not that I'm suggesting for a moment that you personally are spamming, it's the concept introduced here that surprises me).
-------------
And eWhisper,
Of course, the display URLs resolve to 404s, but Google hasn't cared if the display URL does this for a while with subdomains.
That freedom appears to extend to subfolders too, no?
-------------
Just as an aside, in the quest for relevance, when a couple of affiliates for the same merchant touting their direct to merchant product ad suddenly became a no-no, one trend I've noticed is that the competition in many of my segments changed.
Now I try to sell the widget
The ad below me is giving the widget away for "free"
The next ad will pay you to take a widget if you do a survey
The ad following that wants to sell you a list where you can get widgets at incredibly low "wholesale" prices.
I'm not sure if that helps me by being the most honest of the bunch or hurts me by placing me in the midst of bad company.
-----------
With all the changes of late, I can't help but think we're getting away from "relevance" when a machine algorithm tries to determine what's best for the marketplace rather than leaving it up to the market to decide.
Computers can't "read" ads, they can only parse them and sum their components toward some desired score. You know how some search functions provide a relevancy "score" for the results of your query? How often do they hit the mark?
patient2all
If an SEO adage comes into play here, it may be better to use '-' rather than '_' since historically the dash has been taken for a word seperator. Anyone else?
patient2all
I hear you loud and clear. I call it trying to make lemonade out of lemons. I had much higher profits in the old days and am slowly but surely working my way back.
My AdWords bills are much lower. I don't know if this is the case across the board. If it is, it's even worse for G than it is for me!
patient2all
edited for spelling
Is the title, description and displayed url "relevant" so it improves ranking?
Come on, we're waiting for your conclusions!
