Can you be moved into a low AdSense category?

Wondering how a site's category is determined

[guacamolism]

I've been running a site for the last half a year with page views growing steadily from 500 to 8000 per day throughout that time. CPC had been spiky but around 15-40 cents and CTR between 0.5 and 1.5%. Then a couple of months ago I radically improved placement and the CTR jumped to 3.5 - 5%.

Coincidentally, just after I did this Google wrote to me saying I was violating some terms regarding the nature of content and had 3 days to correct or else the account would be closed. Although I’d made the corrections, from that moment (actually just a few days prior to the warning) my CPC fell suddenly then gradually to now 0.03. I’ve tried various things to revitalise it but to no avail. Thanks to the improvement I get 300-400 daily clicks and if my CPC were the same as not so long ago I'd be earning eight times what I’m earning now.

My theory is that during that review, Google had put my site into a low category / topic (e.g. humor / general info) which, as I understand it, means that many advertiser bids do not even enter my pool of eligible ad candidates due to their custom topic filters. So what remains is of lower competition and quality, causing the greatly inferior CPC.

What I’d like to know is whether or not this is a reasonable hypothesis - I mean, is this how the bidding works and has anyone heard of an entire site being manually forced into a specific topic regardless of the actual (keyword) content. Also, is there anything I can do about it (other than changing ad network)?

Thanks for your input.

[martinibuster]

Bill Slawski's in his blog, SEO By the Sea, asked the following question:
[seobythesea.com]

What if Google paid publishers of the search engines’ advertisements more or less based upon a quality score their pages?

He then follows up with a discussion of the patent and how it could be used to pay some publishers more who conformed to high publishing standards and pay less to those who are more or less Panda Bait (my phrasing, not Bills!)

..what if Google looked at other aspects of a web page where Adsense advertisements might be shown, and came up with quality scores for those pages which would determine how much compensation those ads might pay to publishers? A Google patent application... provides some details on how such an approach might work.

Compensation Distribution Using Quality Score

A fascinating read, well worth your time. Full article here. [seobythesea.com]

[creeking]

maybe they used smart pricing

[support.google.com...]

[netmeg]

What I’d like to know is whether or not this is a reasonable hypothesis - I mean, is this how the bidding works and has anyone heard of an entire site being manually forced into a specific topic regardless of the actual (keyword) content. Also, is there anything I can do about it (other than changing ad network)?

First thing I'd probably do in your situation is go look and see how much of my revenue is interest based and how much is contextual.

[guacamolism]

Thanks for those interesting replies - am researching them.

As for the revenue split, it's been quite consistently around 59% contextual and about 38% interest based, plus or minus 4% of variation over the course of time and not much change before and after whatever event occurred.

Not too sure how to interpret those figures.

[Selen]

So Adsense 'ranking score' is used to rank sites in organic search results.. In that case, poor-quality sites that make dimes with Adsense would be better off not having Adsense on their websites at all (because the value of better organic ranking is definitely worth more than some dimes..).

[anefarious1]

I can somewhat relate to what you are talking about. In general, I believe there is an inverse relationship to CTR and CPC although it doesn't seem as extreme as your situation. I cannot explain why this inverse correlation exists but I see it all the time. If anyone knows why please share!

[guacamolism]

The reported symptoms of smart pricing certainly fit the behaviour of my cpc very well so that seems to be a likely candidate. I read that smart pricing status is re-evaluated on a weekly basis so perhaps switching off the ads completely or changing network for a week or so might reset it. Has anyone suspecting having been smart priced ever tried this?

And anefarious, yes I've also noticed that ctr and cpc are very closely and inversely related where even small daily spikes and variations are uncannily mirrored between them. I've always supposed that the extra clicks further deplete the scarcer quality bids that your domain is eligible for thereby lowering the overall average by resorting to more of the lower bids.

Perhaps I will also run a separate test by temporarily de-optimizing unit placement to see if the cpc increases at all. Although not a solution in itself it would provide some interesting information.

One last thing, is a 60:40 split between contextual and interest based revenue a fairly normal ratio?

[netmeg]

I dunno, my ratio is far more skewed to interest based, but most of the contextual ads aren't allowed in AdWords in the first place, so that's probably more a special case.

My impression of smart pricing has always been that it has to do with traffic quality. So if you want to turn the tide back, you might look at where your traffic is coming from, and maybe how you could effect any improvements.

[guacamolism]

Yes that's a very good point. I guess what counts is the clicked-through traffic converting on the advertiser's side once they're on their sites. And I could probably go some way in improving what my site is about from a search result and keyword point of view (which is where the majority of the traffic comes from).

The site has 5.5 million pages though so the re-indexing is going to take some time... But anyway, seems to be the right direction to explore.

[netmeg]

It helps to know who your audience is. When I started out, I thought I knew who my primary audience was, and turned out I was way wrong. So I started writing more to users I had instead of the users I thought I had, and not long after that, traffic and earnings really started taking off. And I have to assume it also had a good effect on ad targeting.

[ember]

I guess what counts is the clicked-through traffic converting on the advertiser's side

That is really all that counts.

[anefarious1]

Ember, are you saying that Adsense knows how well the click through traffic is converting for their advertisers? I didn't think they actually had that information. If this is a factor for publishers, that blows my mind.

[guacamolism]

I believe that this is the exact principle of Smart Pricing - by measuring specific advertiser-defined conversions that the clicked-through user performs (e.g. buying their product or signing up to a newsletter), Google can monitor (and reward) both the relevancy and the quality (i.e. willingness) of the traffic sent there.

[Leosghost]

Ember, are you saying that Adsense knows how well the click through traffic is converting for their advertisers? I didn't think they actually had that information. If this is a factor for publishers, that blows my mind.

Other networks ( Yahoo and MS ) certainly do have this conversion data..and smart price according to conversions..advertisers can see conversion rates ( QS )<= quality score, on a site by site basis for where their ads showed up and got clicked, they can define conversion "goals", They can see which sites send them clicks that dont convert ( meet their defined "goals" ) ..and contact the "ad serving platform" re badly converting sites..These sites then can be "smartpriced" as G calls it based upon their QS..Sites can even be "booted" from the program based upon consistently poor QS..

I'd be very surprised if G were not dong the same kind of QS control as Yahoo and MS do with their "publisher networks"..( publishers on these networks can see their QS on a per channel basis and see changes to QS on a 24 hour basis )..

Netmeg ( who runs adwords and adsense ) may be willing / able to shed more light on how much feedback into QS that G allows advertisers to see and have..

[ember]

Ember, are you saying that Adsense knows how well the click through traffic is converting for their advertisers?

Yes. What is the point of sending traffic to an advertiser's site if the traffic does not perform for him? Why would he continue to pay for that traffic? Adsense/Adwords is all about keeping advertisers happy. They are, after all, the ones paying us. That means sending them quality traffic. Google knows what is quality traffic and what is not.

[netmeg]

Not only does Google know what has converted, Google can predict (or says they can predict) which traffic is likely to convert, and which is not. Heck, I can predict some of it too. If my site targeted to Illinois suddenly gets a ton of traffic from Ubeki-beki-beki-beki-stan, I predict that not very much of it will convert for the advertiser. Likewise, if I'm taking part in some link or click scheme that pays views to register a click on my site, I can confidently predict that not very much of that traffic will convert for the advertiser. Plus I could well lose my AdSense account.

If an ecommerce site is using Google Analytics (and most of them are) Google knows EXACTLY where every sale comes from. I know because I see EXACTLY where every sale comes from in my Google Analytics for client ecommerce sites. And it's not at all difficult to deep dive into which keywords convert, and which display network sites send good converting traffic. The ones that don't convert, I block my ads from showing (or at least my ten and twenty dollar per click ads - I don't mind sending them my three cent ads)

This is what Google says to advertisers about Smart Pricing:

[support.google.com...]

And this is what Google says to publishers about Smart Pricing:

[support.google.com...]

The difference in tone in these two messages is telling.

[guacamolism]

That’s a very interesting perspective and yes quite a different message that G gives to each end of the deal it sets up.

Also fascinating to use different ad networks selectively. I’m interested in what kind of heuristics one could apply to a given page request to determine which network to show. Since their secure search update, Google no longer passes the search terms with HTTP_REFERER, which is a shame because you could filter out users arriving with searches containing words like ‘free’ and ‘download’ for a start. What are the other options? What about geolocating the IP address during the page delivery and showing premium ads to users that are from a list of select (affluent) cities?! That seems kind of ridiculous, but I guess that’s the kind of thing we’re talking about here.. And what about seeing if this is the user’s first page on the site (i.e. just arrived) or if they’ve come from another page on site (indicating higher engagement)? Not sure how well that correlates to quality though.