homepage Welcome to WebmasterWorld Guest from
register, free tools, login, search, pro membership, help, library, announcements, recent posts, open posts,
Become a Pro Member
Visit PubCon.com
Home / Forums Index / Google / Google AdSense
Forum Library, Charter, Moderators: incrediBILL & jatar k & martinibuster

Google AdSense Forum

This 45 message thread spans 2 pages: 45 ( [1] 2 > >     
(not)Smart Pricing - fair play?
this example seems unfair to me

 2:52 am on Jun 12, 2004 (gmt 0)

Okay so I have this page with AdSense, with several months history. Steady low traffic volume, very good CTR and assumed good conversion rate because AdSense payout is high and visitors are very well targeted (low volume, highly relevant visitors, good ads).

I have this friend, who has a site of same niche. He, to, has relevant visitors, steady volume, and good payouts (indicating probably decent conversion rates for Google advertisers).

We meet and get to thinking... what if, on a lark, he were to place a link to my site's comparable page, and I a link to his? What if we did this in an alternating fashion... first he sends me traffic, and then I send him some, and we see what happens to AdSense payouts. Of course we do it in a way that induces click thru to the other's page, sans an AdSense opportunity on the refering page. Of course we stagger it randomly throughout the day, over several days, on both sides.

So my N clicks per M impressions historically paying $X average per click goes to roughly 2N clicks per roughly 2M impressions, and pays what? And his N clicks goes to roughly 2N, while his M clicks go to 2M (roughly)... what do we see with his payout?

I bet you can guess. And when we revert back to normal, without any cross links? I bet you can guess that as well. Stood up to a time test as well.

How can this be fair? How can this be "smart pricing" for anyone but Google?



 3:19 am on Jun 12, 2004 (gmt 0)

I don't understand what you mean.


 3:21 am on Jun 12, 2004 (gmt 0)

I think that he means clicks and impressions doubled but payment remained the same or went down.


 4:20 am on Jun 12, 2004 (gmt 0)

I am not 100% sure i understand your comment.

As a general rule, i would not advise crosslinking between domains. This has lead to google penalties in the past, and maybe your adsense account may be OK, but your SERP standings may fall, or you may totally be removed from the index.

How are you getting double clicks and double impressions if you are simply placing links back and forth? Your traffic would remain the same without another link, so you would be sending away a portion of your traffic to the other site. I do not see how this would be increasing traffic and clicks for both parties unless someone is giving away more traffic than they are getting.

Please explain more. Be careful!


 4:22 am on Jun 12, 2004 (gmt 0)

Yes pls elaborate more cause we have similar problems faced if I undersyand correctly what u mean


 4:42 am on Jun 12, 2004 (gmt 0)

Payment should only increase if there are more conversions.

Smart pricing is just a friendlier way of saying Pay Per Aquisition.

There is only one problem with Pay Per Aquisition and that is you might be able to figure out how someone is converting and game that .. but that's pretty rare and would probably get noticed pretty fast so I wouldn't bother.


 2:29 pm on Jun 12, 2004 (gmt 0)

So... what exactly is the AdSense payment model now?

When it started a year ago it was a simple PPC model: the advertisers paid Google for every click, Google took a cut and paid the rest to the publisher.

Presumably, since the introduction of Smart Pricing, the advertisers still pay Google for every click? So is the amount they pay per click now influenced by the conversion rate (ie. high converting adverts cost more) with the same percentage payout to publishers?

Or is there now a basic PPC payout rate (lower than before) and a PPA bonus for publishers which Google holds on to if the publisher's conversion rate is too low?

It seems like the latter possibility would enable Google to charge the same rates to advertisers as it used to (ie. not have to increase them) while increasing its overall revenue, because it no longer has to pay out as much to publishers whose AdSense panels do not convert.

Is this right? Can somebody who uses both AdWords and AdSense shed some light on this?


 4:00 pm on Jun 12, 2004 (gmt 0)

Lower EPC is passed on to the advertiser. {at least partially, but no reason to believe that fully is not the case}

Working from a small sample here {only a few hundred content clicks since April 1st}.


 9:20 pm on Jun 12, 2004 (gmt 0)

Let's not jump into mythology about cross-linking incurring penalties... or the test periods falling on slightly different sun spot cycles... this was a test and IMHO that "elastic claus" stuff is just a smoke screen for Google to take away earnings from publishers.

Each site was turned off except to pass traffic, so while numbers didn't exactly double they did approximately increase by the expected traffic amounts. As a test, it was good enough to convince all involved that the traffic was indeed transfering sufficient for the test.

As for conversions, sure - another escape route to deny any concerns of bias. These reasonable people do not accept that these two sites, when handling each other's traffic during this test (plus the before and after experiences) would have an impact on conversion rate of any considerbale magnitude.

Both sites saw expected higher impressions, expected higher clicks in proportion to the traffic and prior history of CTR, and *lower* pay per AdSense click on both sides. This was practically uniform through the randomized testing periods as well as afterwards. Both sites are back to normal and back up to the higher pay outs they enjoyed before this little diversion (one is up a bit).

I did the test for a reason. Now I have validated my suspicians that there is a programmed bias that steals from the publishers. I have no interest in figuring out exactly how it works, but I will now work with that understanding. It also clarifies for me why the big media publishers have negotiated payouts (and don't allow "smart pricing"). Who inhis right mind would agree to a deal where the more successful the publisher, the smaller the profit sharing?


 9:32 pm on Jun 12, 2004 (gmt 0)

I think you're jumping into the mythology of anecdotal evidence.

While Google may have implemented smart pricing in a buggy way at this precise moment, it's still a moving target and it's not hard to see that they are converging on pay per aquisition..

The forces that be are pushing them down that path, and even if they haven't arrived at their destination yet, they will eventually have to end up there. Especially when CJ Evolution comes out.

However, I agree and it has been stated before that the fuzziness of the system allows for Google to pick the pocket of the small publisher.

But this isn't a problem with the idea of smart pricing, this is a problem with the implementation. Your comments are interesting, but at the end of the day .. how are they relevant? Different programs pay different in different circumstances. Compare against affiliate links , Ad Sonar, etc etc and whoever pays more plays more.

If your point is that that Google and their compeitition moves towards a pure play per aquisition as soon as possible, then I agree. If your point is that Google has implemented their system to give them lots of room for revenue error .. well, given the lack of competition in this space, would you seriously expect anything else?


 9:44 pm on Jun 12, 2004 (gmt 0)

Also, just a quick question .. Call me stupid, but I had a hard time figuring it out by reading your explanations.

Did the total number of conversions increase? If so why?

My read gave me the impression that you were just recycling total traffic.

If you were you can't expect total revenue to increase.


 11:15 pm on Jun 12, 2004 (gmt 0)

I did the test for a reason. Now I have validated my suspicians that there is a programmed bias that steals from the publishers.

Hmmmm, no you haven't.....


 12:04 am on Jun 13, 2004 (gmt 0)

okay.. I haven't.

So what does this show?

More visitors = more impressions = more clicks on ads = incrementally less revenue.


before 5000 imp 1000 clicks $1000 revenue (average)
during test 10000 imp 2000 clicks $1200 revenue
after 5000 imp 1000 clicks $1050 revenue (average)


 12:13 am on Jun 13, 2004 (gmt 0)

So what does this show?

When you combine two identical sites you will get more repeated clicks due to the same visitors visting both sites. Repeated clicks will be counted as clicks but no revenue.

Did I answer your question.


 12:30 am on Jun 13, 2004 (gmt 0)

I don't get how you have more visitors? I thought you said you 'turned off the traffic'.

Each site was turned off except to pass traffic,

Yer not making sense dude. It sounds like you are recycling visitors, so your total visitors shouldn't increase .. and therefore your revenue should not increase.


 1:33 am on Jun 13, 2004 (gmt 0)

"I don't get how you have more visitors? I thought you said you 'turned off the traffic'."

I think he forwarded the url from one domain to the other.


How long did you forward the url for? There are many posts that talk about how revenue fluctuates as much as 35-40% from one day to the next with the same impressions and ctr.


 1:41 am on Jun 13, 2004 (gmt 0)

CPA, bah. Google makes a mockery of conversion as a reason to reduce publisher's revenues.

Too many AdWords publishers fail to:

a) use Google's conversion tracking scheme
b) use a landing page that encourages conversion

This practice should bear close scrutiny from entities such as the FTC and SEC.


 2:45 am on Jun 13, 2004 (gmt 0)

Just copy your stats into Excel and chart clicks vs. EPC. I have done this in the past (actually I just did it again) and there was no noticeable correlation.


 4:23 am on Jun 13, 2004 (gmt 0)

I don't know why so many who posted misunderstand Paybacksa. This is much closer to experimental design than most other anecdotal evidence posted in this forum. Thanks for sharing it with us.

before 5000 imp 1000 clicks $1000 revenue (average)
during test 10000 imp 2000 clicks $1200 revenue
after 5000 imp 1000 clicks $1050 revenue (average)

Jomaxx, did you mean no correlation in this example, or did you mean on your own site? In this one example, CPC is negatively correlated with total clicks.


 4:40 am on Jun 13, 2004 (gmt 0)

I meant I took 2 months' data from the channel corresponding to my main AdSense site and saw no particular pattern.

Frankly there are so many variables that have to be accounted for that I probably wouldn't accept any such test as proof unless I designed the test and analyzed raw the numbers myself. But I just don't see the point in it. Google's algorithm is a black box, and you can either stay in the program if the numbers are good enough, or quit the program if you can find something better to replace it.


 5:44 am on Jun 13, 2004 (gmt 0)

Agreed. I'm still in. Here's why I think this exercise is relevant: one must decide how much to invest in generating traffic.

While it's understood that EPC varies with time, I expected either zero correlation or a small but slightly positive correlation between total Adsense clicks and EPC for any given site, all else constant. Call it the default hypothesis (H0). It is pretty much taken for granted.

Now, suppose I want to invest money to increase traffic. Also, suppose Adsense is the main revenue generator on the site.

If H0 were true, then I could increase my marketing budget to generate incrementally more traffic for the site without fear of decreasing ROI.

If the alternate hypothesis were true (it is possible that there is a negative correlation between total Adsense clicks and EPC), then I would consider NOT increasing my marketing budget for this site -- or at least do so conservatively.

I was surprised by paybacksa's report, but it suggests the possibility of a negative correlation under certain (unknown) circumstances. Of course it does not prove it.


 6:58 am on Jun 13, 2004 (gmt 0)

I think the misunderstanding is that impressions and clicks are meaningful to smart pricing.

It's about visitors and conversions. Getting 100 visitors to click 10,000 times versus 100 times should not necessarily change your revenue.


 7:27 am on Jun 13, 2004 (gmt 0)

I have been keeping a daily log of a particular channel, in regards to the revenue it generates, the unique visitor traffic it receives, the cost to bring in that traffic (for that portion that is through natural search), CTRs, ECPCs, the ads being displayed, etc.

With this channel, I can see my results roughly coinciding w/ Paybacksa's test. Over the last month or so, I have been able to triple unique visitor traffic to this channel. Impressions are up ~3x, clicks are up ~3x, CTR has remained pretty constant. Revenues are up about 12% since the beginning of the month, despite the ~3x increase in unique traffic and clicks.

I have seen this anecdotally on other channels as well, and am now watching it more carefully to see if the pattern continues.


 8:04 am on Jun 13, 2004 (gmt 0)

Don't forget Heisenberg. When I was in college I watched the jai-alai results alot and noticed that the total return from the two teams/players that placed was always greater than sixteen dollars, with eight players/teams per game. I watched this through the paper over weeks and it happened without exception. I thought I had found a flaw in the program and I could exploit it. I went to the orlando jai-alai and bet ($2) all eight teams to place in every match one night. Guess what? It never paid $16 the entire night. Heisensberg principle? No, just parimutual betting. So whats going on with Google? Is it Heisenberg? Is it like paramutual betting? Is there an inverse relation between ctr and epc?
I suspect a little of all three.

[edited by: Powdork at 8:10 am (utc) on June 13, 2004]


 8:08 am on Jun 13, 2004 (gmt 0)

Yeah, ok, now I'm really confused.


 8:27 am on Jun 13, 2004 (gmt 0)

Its simple. In parimutual betting, all the bets are taken into consideration to determine the payout for any given bet. When I was reading the paper it was obvious that I could make 8 $2 bets and get back more than $16. Making the 8 $2 bets, however, changes the payout ratio so that I was betting $16 to get back something like $15.50. Heisenberg's principal (simplified) is that by observing something, you have an impact which affects the outcome of that which you are observing.
Basically what I am saying is that there are a lot of variables going into dumb pricing, many of which could have been altered unknowingly with the experiment. I'm also saying that Google uses this indeciperability to lower publisher payouts in the name of smart pricing. They're still giving us a good deal in many cases, we were just spoiled by their introductory rates, which were a bit unrealistic in many cases.
They are no longer head and shoulders above the rest, but the rest are yet to arrive.


 8:42 am on Jun 13, 2004 (gmt 0)

Got it. I suppose I should have just googled parimutual betting.

But why bother when you have such accomodating WebmasterWorlders... ;)


 2:29 pm on Jun 13, 2004 (gmt 0)

Yes, experimental biases are often a problem. This is why it is important to "control" for external factors.

There are well known generic experimental designs used by researchers and academics, and guess what.... In most controlled experiments, some form of bias is present.

Paybacksa's method reduced some biases by alternating the experimental treatment between the two websites. It would have been more convincing if instead of "switching off" one of the sites, one could have one site run with its normal amount of traffic as a control. (Perhaps a third website would have been useful here).


 4:00 pm on Jun 13, 2004 (gmt 0)

BTW, I'm not saying some phenomenon is not there, but the data does not support paybacksa's rant about how Google "steals from the publishers". For example...

- Maybe the traffic from the two sites isn't identical, and people who get funneled from one site don't tend to click on the other site's most expensive ads. Frankly I'm surprised two sites could be so interchangeable that you could seamlessly redirect traffic from one to another without consequences.

- Maybe the matching algorithm somehow tends to show broader, or at least different, ads as the number of impressions goes up. I doubt one site is blowing any advertiser's entire budget, but maybe AdSense portions out clicks so that not too much of an advertiser's content traffic comes from a single site.

- Maybe when there is a sudden spike in traffic, the extra clicks are charged to advertisers at the lower rate. Then if traffic stays at a high level and the "quality" of the higher level of traffic is established, rates would go up.


 4:08 pm on Jun 13, 2004 (gmt 0)

CPA, bah. Google makes a mockery of conversion as a reason to reduce publisher's revenues.
Too many AdWords publishers fail to:

a) use Google's conversion tracking scheme
b) use a landing page that encourages conversion

I agree on (a) and I've written about this since the first days I joined Adsense as a publisher.

"Smart pricing" can't be very smart without some measure of effectiveness in place at the advertiser's pages.

Currently, in absense of conversion data, Google's "smart pricing" seems to give the benefit of the doubt to the advertiser.

So, looking at my logs, a visitor from United Arab Emirates on a page about oil refinery, clicks on an ad of Shell on improving refinery plant performace and on another ad on oil pipe hangers and supports. And those clicks earn a few cents.

I think that for advertisers who don't use conversion tracking, the benefit of the doubt should be given to publishers.

This 45 message thread spans 2 pages: 45 ( [1] 2 > >
Global Options:
 top home search open messages active posts  

Home / Forums Index / Google / Google AdSense
rss feed

All trademarks and copyrights held by respective owners. Member comments are owned by the poster.
Home ¦ Free Tools ¦ Terms of Service ¦ Privacy Policy ¦ Report Problem ¦ About ¦ Library ¦ Newsletter
WebmasterWorld is a Developer Shed Community owned by Jim Boykin.
© Webmaster World 1996-2014 all rights reserved