in: [webmasterworld.com...]
AWA wrote:
It is still Max CPC x CTR. The difference is that now, the CTR of the ad copy itself is factored in, instead of it being solely the CTR of the keyword.
Would that be the CTR of the ad in combination with each specific keyword?
And how does this feature compare to "Automatically optimize ad serving for my ads" at campaign level?
Thanks in advance for explanations!
Talking about the AdRank formula:
max CPC x CTR vs. max CPC x CTR x RAD
Honestly, I think we'd do well to let 'max CPC x CTR x RAD' sink into oblivion soon, before it causes a lot of confusion. Or, at the very least, let's always identify it as pure speculation. And if you mention it to AdWords support folks, please don't expect that they'll have any idea as to what you're referring to. ;)
Thanks in advance for explanations!
I'm sorry to disappoint, but dedicated 'algo hunters' will know that nitty gritty details are not routinely released. I did my best to set this expectation, as I recall, with my first post on the subject, many weeks ago.
That said, I'll certainly check with the folks who are more in the know than I, to see if there's anything that I can add to what I've already said.
AWA
One reason I asked is because in some groups with multiple ads, some ads are barely showing, while I've switched optimisation off. So I wondered what the difference between optimisation on and off is. The new meaning of CTR could explain this behavior.
One reason I asked is because in some groups with multiple ads, some ads are barely showing, while I've switched optimisation off. So I wondered what the difference between optimisation on and off is. The new meaning of CTR could explain this behavior.
Actually this is a bug on Google's side. A temporary solution is here [webmasterworld.com...]
We paid thousands of extra dollars (we think, as much as we can tell) because of this unannounced change in pricing logic.
Keeping us in the total dark has damaged us financially.
Can AWA or anyone else explain the necessity of not informing advertisters?
I invite other forum readers to call Tech support and see if you can get a straight answer on this, even now.
But such a practice cannot be carried over to the Adwords system. For after all we do pay money for each and every click and Google does require better communication so that we can optimally spend our money. One just has to look at things like the recently introduced budget optimizer, to understand that Google is not always correct.
I am not saying that there is anything wrong with this change. To the contrary I have been able to write much more relevant ads and seen a small drop in my CPCs. But it still stands to reason that, we advertisers be kept in the know.
My problem and criticism are simple. If the goal post is being moved, please tell me of that fact
If we had known that changing the text of our ad might increase CPC's because of a lack of CTR history for the ad, then we would not have changed 21 ads all at the same time.
The change has been effective CTR is up 15% (currently) but costs are way up as well.
I suspect that the change in ranking and pricing formula is part and parcel of the Budget Optimizer.
I have to say that I am a little taken aback with some of the posts in the two threads on this subject, and will do my best to address the concerns expressed.
But I have to wonder, has anyone ever not known that one's ad copy has always been inextricably linked to one's CTR? And really, how could it be otherwise.
A silly example to illustrate the point:
If the keyword is 'organic puppy chow' and the ad text talks about Organic Puppy Chow, the ad/keyword will likely have a pretty good CTR, which tends to improves position, and may lower CPC at the same time. Nothing new here.
On the other hand if the keyword is 'organic puppy chow', and the ad text talks about Corgi Model Cars, then the ad/keyword will likely have a pretty bad CTR, which tends to lower position, and may raise CPC if one wants to maintain position. Nothing new here.
And for the keyword 'organic puppy chow', if one changed their ad text from the first example above to the second, then the CTR, position, and CPC are all likely to suffer. Nothing new here either, right?
We paid thousands of extra dollars (we think, as much as we can tell) because of this unannounced change in pricing logic.Keeping us in the total dark has damaged us financially.
My problem and criticism are simple. If the goal post is being moved, please tell me of that fact
As mentioned in another thread, ongoing ads-quality changes were first announced in July 2004. And the importance of ad text to position has been a mentioned in the FAQ since January, as for example on this page:
How are ads ranked?
[adwords.google.com...]
Quoting from that page:
Your ad is ranked on the search results and content pages based on various performance factors including: maximum cost-per-click (CPC), clickthrough rate (CTR), and ad text. Having relevant ad text, a high CPC and a strong CTR will result in a higher position for your ad...
I am not saying that there is anything wrong with this change. To the contrary I have been able to write much more relevant ads and seen a small drop in my CPCs. But it still stands to reason that, we advertisers be kept in the know.
In a sense, this post nails it. A very important key to AdWords success is (and has always been) to write ad text that is carefully targeted to the keywords. It simply makes good sense to show a user an ad for the same thing that they're searching for. If an advertiser doesn't do this, well, typically, the results will be less than stellar.
If we had known that changing the text of our ad might increase CPC's because of a lack of CTR history for the ad, then we would not have changed 21 ads all at the same time.
Honestly, I'd say that 'lack of history' is much less important than simple targeting.
Whew! Sorry for the very long post!
AWA
Prior to the change. Ranking was based on CTRxCPC with no additional mathematical varriables. Correct?
The CTR was on a word by word basis.
So If I have a word with a 10% CTR and make a minor tweak in my ad it is unlikely to change the CTR for the word and thus unlikely to change my pricing per click or position.
Of course, I know that if I word the new ad poorly over time[i] the CTR of the word might change for the worse.
But now (correct me if I am wrong) The ad's CTR is also "factored-in" in some way, mathematically - directly. (I would like a clear cut answer on this)
So if I change the ad then [i]instantly my CPC will increase for the same placment because the new ad has poor CTR because it has NO history.
It seems the the CTR "recovery" takes some time and the short term "recovery" of the new ad is not weighted the same as long term history.
The result is that changing an "if" to a "but" in your ad can significantly increase your spend until the new ad's good CTR "proves" itself. So in the short term your cost increases for no other reason than the algo. In our case it is (to us) alot of money.
I can't be sure how much because I can't get a straight answer from anyone.
Please keep in mind that our ads have HIGHER ctr than the old ads but our costs have dramatically increased.
This has caused us to NOT try to improve our ads, which can't be what Google wants.
Can't agree more. There is this fear of editing our ads. Even if we want to change something in the ad, we'll hold back. The original ad can be misleading to surfers but what can we do?
Is there a way to "regain" our CTR after each review? It is crazy to start all over again (paying much more to get a reasonable ad rank) just because of a small change in the ad text.
And this is not to say change URL.
Please help!
In our situation it cost us roughly an extra $1k per day.
How long do you think I'm going to be willing to test a new ad, and how many new ads do you think I'm going to want to test, when they are costing me an average of $50 an hour each just to have them running?
This is a screen shot of 2 charts provided in adwords reports.
<url removed>
It's Average cpc and Cost. (stipped numbers and names out, only the graph). Imagine your own numbers in that chart. Anyone's accounting department would go ballistic if they saw this. And I haven't been doing anything at all different that what I've been doing since we started using adwords.
I think it looks like a big money grab on Google's part, although as I said, I completely understand the need to make improvements. But why at our expense? And why NO WARNING.
grrrrrrr! :-(
[edited by: eWhisper at 4:04 am (utc) on April 20, 2005]
[edit reason] Please No URL Drops See Tos [webmasterworld.com] [/edit]
I'm still trying to hunt down the folks I'll need to talk to, to see if I can get additional algo details.
Are you still working on this one AWA? I'd very much appeciate that.
On the other hand if the keyword is 'organic puppy chow', and the ad text talks about Corgi Model Cars,
I hope it's clear that most advertisers are wiser than that, especially professional advertisers (hopefully). But our AdGroups may contain several synonymous words for the same subject, as well as our ads. Sometimes they're not yet worth their own adgroup when you start out. Or one ad may advertise one USP, while another ad may mention another. One ad may work better for keyword A and the other for keyword B. Thats why I hope keyword/ad-combinaion is taken into account, and not just ad CTR in general.
If ad CTR is taken into account, one way or another, shouldn't this topic [services.google.com] in the learning center be updated?
Why can't we get a straight answer on this?
We have spent a considerable amount of money in much higher CPC as a result of this change in rank formula/pricing.
Are you still working on this one AWA? I'd very much appeciate that.
I've done a fair amount of asking around HitProf, and have to report that I'm not able to detail the algo any further than I've already done. As business owners, I hope that you'll understand that it is not to a business's competitive advantage to be too precise about exactly what they do, in a very public forum.
From the advertiser perspective, I'm fully aware that more information is better than less information - and we're working on ways to better keep our advertisers informed. (Do keep in mind, though, that this has to happen in something like 41 languages. ;))
So, back to the algo. I've said before Google is very concerned with delivering relevant ads to our users - and, as was announced back in July 2004, there have been (and will continue to be) an ongoing series of incremental changes designed to improve our ability to do this. The algo change that incorporates info about the ad copy itself is a part of this. And, over time, I think it'd be safe to say that more factors, rather than fewer, will be evaluated as a measure of relevance.
One ad may work better for keyword A and the other for keyword B. Thats why I hope keyword/ad-combinaion is taken into account, and not just ad CTR in general.
I can confirm that is the combination of the ad, and the keyword that brings it up, that is taken into account.
At this point I would, myself, settle for even a straightforward admission that ad CTR is mathematically factored it. The phone support folks still categorically say it is not.
I almost wonder if this is a confusion revolving around wording, nyet. I'm not a mathematician, but it's my take on the question that if relevance factors concerning the ad copy are a part of the algorithm, then, yes, it is "mathematically" factored in.
I'll follow up with a phone team manager to make sure we're all on the same page. I've already done a spot check, though, and everyone seems pretty clear that ad copy can, and does, influence position. And a lower position combined with a high Max CPC can result in an increased average CPC, as the system works to maintain one's position.
We have spent a considerable amount of money in much higher CPC as a result of this change in rank formula/pricing.
Nyet, this actually makes me a little uneasy. You are describing a very direct cause and effect relationship here as if it were a fact. It may be related, as described briefly above, but judging by the time frame you've mentioned in other treads, I am not sure this is the only factor. If I recall correctly you described an increase in CPC about six or seven weeks ago, which means in early March - while the algo change was made much earlier, and was first discussed here in January.
Lastly, please know that I've been forwarding your feedback, from the several threads on this subject, to the related teams, all along.
AWA
The CPC issue for us was about 10 weeks ago. It was noticed within 24 hours of changing the text on *all* of our ads to be more directly competitive with out competition.
Instantly our average CPC went up considerably.
Not knowing that changing one's ad text would *directly-mathematically* affect pricing we thought we had a bidding war on our hands.
After a few weeks with the "bidding war" still going (we thought) we also noticed that although the new ads were performing nicely with higher CTR they also mysteriously had higher CPC.
A certain amount of "noise" can be expected, but this was across the board, every ad, every ad group, the whole campaign.
*Exactly* what on would expect if AD CTR is being mathematically factored in and new ads mean you start at zero.
So, we surmise that had we *known* of this change we would not have changed ALL of our ads at once, but would have done some testing.
So it is actually a failure of language. And just like in your post above, to say the "changes will be made to increase ad relevancy...." IMHO is not sufficient warning that a tried and true, well known, oft discussed, central pillar of adwords operation, pricing and ranking will *fundamentally* change and should you not make the correct inferrence you will *lose* a lot of money needlessly.
So while I very much appreciate your response. I still think this change could have been handled *much* better by Google.
Here is a statement the give away nothing about the Algo but would have prevented the loss of money from us.
"Google will be making a slight alteration in the ranking and CPC pricing formula which will mathematically take into account ad CTR. This may effect pricing of new ads relative to their initial unproven CTR history."
It is a fundamental change so IMHO it should have gone out in an email.
See?
Can you direct me to the notification you mention?
We are out real money here and none too happy about it. We appreciate trade secrets but don't think the "goal posts" should get moved without notification.
And because the CPC of the new ads has slowly been equalizing with the old ads (hopefully it will go even lower since our CTR is higher on the new ads). I am at a loss as to how to interpret this *other* than to assume that eventually the CTR history of the new ads will be "factored-in" enough. Again, exactly what one would suspect if ad CTR is mathematically factored in.
I am no mathematician or statistician, but the divergence of CTR and CPC over this length of time cannot be "noise".
So the extra money we have been paying is either a bidding war (a real one) or an "Algo Tax" because of ad CTR. If it is a bidding war, what are the old ads still getting the OLD CPC?
I am sorry but the "evidence" seems to point to "algo tax".
I am open to suggestion otherwise.
I can confirm that is the combination of the ad, and the keyword that brings it up, that is taken into account.
This is absolutely not true.
If it is, can you please explain to me how it is possible that I create a new campaign, create a brand new ad, throw in some keywords, and they are instantly disabled with 0 impressions?
And yes, these are keywords that I have NEVER bid on before.
And a lower position combined with a high Max CPC can result in an increased average CPC, as the system works to maintain one's position.
Uhm? Don't understand this (could be due to a lack of coffee though).
AWA, it is extremely critical that we have some idea into how the CPC is calculated. Let us take the case below where 2 ads getting equal number of clicks.
Ad---CTR---Income---Expense---Profit
**************************************
A----5%----100------120------(20)
B----1%----200------120------80
**************************************
Overall 3% 300 240 60
It might be easy to conclude that Ad B is better. This would pretty much be the profitability if the CTR of the Ad is already taken into account in the CPC.
In the case of only Keyword CTR being used, our assumption was that only the overall CTR was used for the CPC calculation. We used to adjust the CPC for each ad to take into account how the CTR of a particular ad affects the over all CTR. So for Ad A in the absence of Ad B the CPC would have reduced as the overall CTR would have increased to 5% from 3%
Ad---CTR---Income---Expense---Profit---Adj.Expense---Adj.Profit
*****************************************************************
A----5%----100------120-------(20)-----120*3/5 = 72---28
B----1%----200------120--------80------120*3/1 = 360--(160)
*****************************************************************
Taking this into account, we actually see that Ad A is better.
As you can see it makes a lot of difference for us to know what the formula is. Atleast some idea about it.
[edited by: inasisi at 8:26 pm (utc) on April 26, 2005]
However, we absolutely *must* know *what* things directly affect pricing.
I cannot find any disclosure on the adwords site that ad ctr is mathematically factored into rank and price.
We all know it has *always* been related because better ads get better CTR *over time* for the words involved.
But never before did changing the ad copy *immediately* and in a direct way affect CPC.
I can't make out what you are saying.
AWA reacted (in that same thread) on this:
This sort of illustrates the problems inherent in publishing a speculative and unverifiable "formula" in a public forum, as if it were a fact.
It is not quite fair, because I did not publish the formula "as if it were a fact", but I have written in the 102 msg this: "I HAVE NEVER SEEN this formula ofically, it is how I understand the old info + AWA info.
So, no fact, but HOW I UNDERSTAND.
I also understand that "my formula" maked the phones ring and gave some hard time to support people and maybe to AWA too.
Sorry for that! I very appreciate your involvement and your PATIENCE with all of us here :-)
But this is so important issue, that we really need to understand it.
I REALLY do not care if the real formula is:
max CPC x (CTR of KW) x (CTR of ad)
or
max CPC x (CTR of KW) + (CTR of ad))
or
max CPC x (CTR of KW) + (CTR of ad)/3,14)
or
CPC x ((CTR of KW) + 2x(CTR of ad))
or anything else.
PLEASE DO NOT ASK THE SUPPORT WHICH OF THEM IS THE RIGHT ONE. These are just MY exemples to illustrate that I DO NOT CARE what the real formula is.
The fact is, that the CTR of ad IS really in the game. It is enough for me.
The key question was asked by NYET and my version is this:
May the history of ad text CTR influence the AdRank?
This is not a nitty gritty technical question, but the question about the price of the product. It is only fair, that the customers know ALL the factors that may influence the price.
If the answer to the question is YES, then, if we touch the ad text, we have to take this into consideration.
If the answer NO, then we know that we may touch the ad text and do not need worry about its history.
What I meant to say is that depending on the formula used (Keyword CTR or Keyword & Ad CTR), deciding which Ad has the better ROI, can vary significantly. So it is critical that we have some ideas about the calculations.
So it is critical that we have some ideas about the calculations.
No. We do not need to know exact calculations.
The first thing is to have a good ad text. If the text is good you do not need to care how they calculate its rank.
Lets assume you know the exact formula. How it will help you to get the higher rank? Only by writting better ad text! You can do this without knowing the formula.
