Welcome to WebmasterWorld Guest from 3.80.4.76

Forum Moderators: martinibuster

Message Too Old, No Replies

A way smartpricing could be implemented

(a purely speculative and useless thread)

     
3:15 pm on Jan 7, 2007 (gmt 0)

Preferred Member

10+ Year Member

joined:Feb 21, 2005
posts:553
votes: 0


The nature of smartpricing has been long discussed, and the official voice of Google on the matter is that there are several factors in it, the most important being the conversion rate of the clicks...

Often, the look of it is that it's quite random...

I just had an idea of a simple way Google could have to determine which clicks should be smartprices mre or less: the ratio of "quick bounces".

I mean: if the user clicks on an ad but after 3 seconds is back to the publisher, this is certainly a useless click for the advertiser.

Count how many "quick bounces" each site has, and you might have an index on how much the site should be smartpriced.

This is It's just an idea that I had: as I said, it's pure speculation, plus it's completely useless: those are factors you cannot control.

3:52 pm on Jan 7, 2007 (gmt 0)

Senior Member

joined:Oct 27, 2001
posts:10210
votes: 0


The nature of smartpricing has been long discussed, and the official voice of Google on the matter is that there are several factors in it, the most important being the conversion rate of the clicks...

Actually, it's the likelihood of conversion, as determined not just by conversion tracking (which isn't used by all advertisers), but by other factors such as type of content on the page (which Google has mentioned specifically in its explanation of smart pricing).

For all we know, "bounce rate" could be a factor in smart pricing, though it isn't likely to be based solely on how quickly a user hits the back button (since, if the referring page was worthless, the user may have clicked on the ad to escape the worthless page). It is worth noting that "bounce rate" is one of the stats provided by Google Analytics.

4:22 pm on Jan 7, 2007 (gmt 0)

Preferred Member

10+ Year Member

joined:Aug 6, 2005
posts:575
votes: 0


Correct me if I'm wrong, but Google (nor the advertiser) has any way of knowing when the user hits the 'Back' button. OK the advertiser can see from his logs that the visitor landed but then browsed no further, but they don't know how long the visitor stayed on the landing page for.

As far as I am aware, webmasters and (stats software) can only 'guess' when a visitor leaves a website by tracking which pages the IP address visited. The only other way that I know of that an advertiser could see when the visitor left would be to have an 'onUnload' JavaScript in the page body. And as far as I have seen, Google doesn't use anything like this.

I suppose that Google could see how long it was before they served another ad block to the same visitor on the publisher's website and 'predict' how long the visitor had stayed on the advertiser's website. But then what about proxies and AOL? People could be smartpriced erroneously if this was how Google determined our fate.

Just my 2 pence. :)

4:55 pm on Jan 7, 2007 (gmt 0)

Senior Member

WebmasterWorld Senior Member 10+ Year Member

joined:Apr 26, 2006
posts:1397
votes: 0


Nothing wrong with speculation on this subject. In fact, when there's so little from Google, it's one of few available ways to figure out Smart Pricing, and that can be practical for your own website(s). A lot of stuff starts to make sense when you ask yourself, "If I was Google, what would I do?"

I think bounce rate is a likely factor. If not one of the biggest issues, certainly high on the list. Although it essentially amounts to the same issue as Page CTR. But I think Google should penalize the advertiser as much as the publisher for a high bounce rate, if it's going to value this issue. I mean, if many people click on your ad, and then quickly move on to another ad, at some point you have to wonder if it's the ad or advertiser's site that has the problem, not the placement of the ad on the publisher's site.

Online advertising in the early years was about conversion (affiliate marketing, e.g., Amazon, etc.). That of course is so easy to track and it has few fraud concerns, because it is so objective. Ads sold based on clicks instead of conversions will always be a tricky issue, especially when there's no sale involved on the target site, because it's so subjective.

I'd like to see some kind of 'bad ad/pub' reporting system where Adsense and Adwords users can 'flag' each other, but only directly to Google. As the complaints stack up, then they can weed out abusers based on something more objective than whatever they do now, some kind of algo.

I'm starting to believe Google has changed its Smart Pricing from account-wide to site-wide, and even page-only. Just when I thought I'd been smart-priced, I get a click over $2; whereas another page is continuously only $0.02 ea.

Maybe advertisers will one day be able to bid on pages instead of sites or accounts.

p/g

6:19 pm on Jan 7, 2007 (gmt 0)

Full Member

10+ Year Member

joined:Nov 26, 2006
posts:292
votes: 0


I'm starting to believe Google has changed its Smart Pricing from account-wide to site-wide, and even page-only. Just when I thought I'd been smart-priced, I get a click over $2; whereas another page is continuously only $0.02 ea.

I agree with you my newer sites get 0.2 to 0.5 per click. My older site and forum some time jump upto $5.

6:31 pm on Jan 7, 2007 (gmt 0)

Senior Member

WebmasterWorld Senior Member 10+ Year Member

joined:Mar 1, 2004
posts:1258
votes: 0


I'm starting to believe Google has changed its Smart Pricing from account-wide to site-wide, and even page-only.

I never thought it was anything more than "page-only". More than that simply never made logical sense to me.

I think every site probably has one or more "stinker" pages that just aren't everything that they could be. It would make sense for those pages to not perform well, while other site pages are performing up to and perhaps beyond expectations. Finding and fixing those "stinker" pages may or may not be easy. It all depends...

I think we've all seen search results that give a ranking of how closely each result matches what was searched for in a percentage type display (92%, 83%, 58%, etc.). I think that sort of relationship matching may well play a part although a search is not involved.

6:37 pm on Jan 7, 2007 (gmt 0)

Senior Member

joined:Oct 27, 2001
posts:10210
votes: 0


I'm starting to believe Google has changed its Smart Pricing from account-wide to site-wide, and even page-only.

It's only reasonable to think that smart pricing has evolved since it was introduced back in April, 2004. Google has had time to gather and analyze a lot of data in the last 2-3/4 years.

1:33 am on Jan 8, 2007 (gmt 0)

Senior Member

WebmasterWorld Senior Member 10+ Year Member

joined:Apr 18, 2005
posts:817
votes: 0


I mean: if the user clicks on an ad but after 3 seconds is back to the publisher, this is certainly a useless click for the advertiser.

What if the user clicks on an ad, opens another window with that result, and uses one window to go back to the publisher's site? Then the publisher is being unfairly smartpriced.

8:18 am on Jan 8, 2007 (gmt 0)

Senior Member

WebmasterWorld Senior Member 10+ Year Member

joined:Aug 11, 2006
posts:902
votes: 0


Who said anything about Smartpricing having to be fair?

Obviously it would be good if it were to be fair, but G may consider it acceptable it if got it right most of the time.

7:04 am on Jan 11, 2007 (gmt 0)

Senior Member

WebmasterWorld Senior Member 10+ Year Member

joined:Apr 18, 2005
posts:817
votes: 0


The value of a publisher's site to an advertiser is conversions/ROI, not time spent. So any unfair smartpricing would act as a disincentive for the publisher to display those ads.
7:24 am on Jan 11, 2007 (gmt 0)

Preferred Member

10+ Year Member

joined:Feb 21, 2005
posts:553
votes: 0


The value of a publisher's site to an advertiser is conversions/ROI, not time spent.

Sure, but G having very little information about conversions they will have to approximate that somehow. Time spent could be just one factor of getting this "approximation".

1:40 am on Jan 12, 2007 (gmt 0)

Senior Member

WebmasterWorld Senior Member 10+ Year Member

joined:Apr 18, 2005
posts:817
votes: 0


Time spent on a website is a slippery concept at best. Rather than go down a path where even approximate figures are difficult to get (and fraud is easy to commit), the engines are much better off implementing smartpricing based on the limited information they do have that suggests poor conversions or ROI.
2:38 am on Jan 12, 2007 (gmt 0)

Senior Member

WebmasterWorld Senior Member 10+ Year Member

joined:July 1, 2002
posts:1427
votes: 2


They could use conversion data that they have from a subset of advertisers that use google tracking. After that, they could train a neural network or naive bayesian classifier one that data. Then, they could use that classifier to estimate the likelihood of conversion for publisher's page/advertiser's landing page pairs across the board.

But that's just speculation. There are plenty of ways to do it when you have unlimited brain power and hardware resources at your disposal.

 

Join The Conversation

Moderators and Top Contributors

Hot Threads This Week

Featured Threads

Free SEO Tools

Hire Expert Members