Advertiser’s smart filtering

The alternative reason for smart pricing effects

[lammert]

In a recent thread [webmasterworld.com] about falling earnings in AdSense I mentioned that Advertiser behaviour can be a reason that earnings get lower after an AdSense site is on-line for some weeks.

I am both an AdSense publisher and AdWords advertiser and therefore have experience from both sides of the line. In this post I will explain how I use the AdWords site filter and negative keyword options to remove my advertisements from selected sites.

Two types of campaigns

I run two types of campaigns. The first type are low price, broad audience campaigns. I use these campaigns to get as much as possible visitors to a site. Conversion is not an issue. The main reason for these campaigns is to get known.

Recently I have run such a campaign for example to promote a forum about widgets. With general keywords related to these widges I was able to get my ads on many publishers sites.

I do not filter sites or negative keywords with this type of campaign. Why should I? Clicks don't cost much--often less than $0.05--and you never know at forehand which visitor will be interested and which one not. The main selection is already done with the ad copy.

The second campaign type I run is to sell specific widget types. These widgets will be bought by a very narrow group of potential customers. To make my ads visible at spot #1 or #2, I have to bid quite high in AdWords. Publishers with high quality sites often use only one small ad block, for example a 468x60 banner. To get my ads in that ad block, I also have to bid high in the content network. The negative effect of this is that my ads will also be in #1 or #2 position on all the MFA and off-topic sites which match my keywords. Therefore I daily scan my server log-files with an AWK script to extract the AdSense referrers. I go to the sites and decide if they are worth my AdWords money. If not, I either put the site in my AdWords URL filter, or add negative keywords to my campaign which are specific for the site.

Reasons for filtering publishers

The reasons I decide to stop advertizing on a specific publisher's site are diverse. Sales are not high enough in my niches that I can solely rely on the statistics from AdWords. I therefore use my own set of rules to decide if I want to filter or not.

• Password protected sites. I often read on this board questions about displaying ads on password protected sites, mainly forums. If I get a referral from a site and it happens to be a password protected site I always put that password protected site in my AdWords URL filter. Why? Simple, I don't know what is talked about on those pages, maybe it is on topic, maybe not, but I certainly don't spend my money on a site which doesn't allow me to view its contents.
  
• Slap the ad in your face sites. Is the viewable part of a page on my 1024x768 screen fully filled with AdSense and/or other advertisements? Good reason for me to filter the site. People that visit such a page don't select an ad because of the content of the page, they click because it is the only thing they see. Although I may loose potential customers, these sites help create a negative image of the internet, and I don't want to sponsor that.
  
• Negative comments about widgets. If I want to sell widgets and a site is having bad information about these widgets, chances are low that an interested visitor will ever buy a widget from an ad on such a site. I often use negative keywords for this: keywords that stop my advertisements showing on sites where those keywords are mentioned.
  
• Different regions. The widgets I sell are region related. There are really good informational publishers about my widgets, but some of them are talking about these widgets in other regions on earth. A visitor to such a site would almost never be interested in my widgets. Therefore I use names of competing regions as negative keywords. Please understand that this is something different from regional advertizing. If you have a restaurant in a city, you wouldn't be interested in people searching for a restaurant from an IP address 1000 miles away. That is when regional advertising helps. In my business, the same types of widgets are offered from different regions. Someone from far might be very well interested in my widgets, eventhough his IP address is 1000 or more miles away. Therefore regional advertising doesn't help in my case.
  
• Demographic reasons. My widgets are interesting for a specific group of people. People with specific interests, background and age. Sometimes my ads appear on sites which are on-topic and with a high pageview number, but where the audience is different from my wanted audience. In my case the widgets are specifically interesting for somewhat older people. When I see my ads appearing on sites for younger people (often determined by subjects like gaming, music groups etc) I often decide to put the site in my URL filter. This is many times not a negative vote for such a site. Many of these sites are good for AdSense ads, they are just not good for MY ads.
  

The effects on the AdSense publisher

I expect to be not the only AdWords advertiser who checks his referrals from the content network in such a way. I am therefore not surprised that many publishers mention here that just after launching a site, earnings are good, but that after some days or weeks EPC is falling.

Smart pricing is often mentioned as the reason, but why would Google deliberately lower earnings for a site so short after it went life? In my opinion there are two ways smart pricing can work. One way is by analyzing the page content and calculate a conversion factor. This can be done immediately when first scanning the page with the Mediabot.

If on the other hand Google uses the conversion tracking reports of advertisers to determine of smart pricing should kick in, one or two weeks of statistical data is often not enough to determine the smart pricing factor. I would expect the smart pricing effect to take longer than one or two weeks when it is based on conversion tracking statistics.

Therefore if your site went life and earnings collapsed after some days or weeks, look carefully at the ads that display. Are the same ads still present? Only one or two high paying ads which are retracted from a site can make a huge difference in average EPC when the remaining ads are in the $0.05 group.

If publishers or advertisers have experience with this, please post them in this thread so we can try to determine if it is Google, or actually the advertisers who cause EPC falling for some publishers' sites.

More on the AdWords site filter in this thread [webmasterworld.com].

<edit>Speeling</edit>

[Hobbs]

Excellent insight lammert, thank you very much, publishers often miss the advertiser's perspective, the Adwords forum should be compulsory reading for AdSensers.

[21_blue]

lammert, very interesting, thanks for posting.

You haven't mentioned site-targeting (CPM ads). Do you use that and, if not, is there a reason?

[lammert]

I don't use site targetted CPM ads. For my type of widgets and niche they are not appropriate.

[ken_b]

Lammert; Thanks for the insight into your thinking. I can sure see how publishers might blame the results on Smart Pricing.

[lammert]

Smart pricing is something I assume to kick in right away, after an algorithmical check of the page content, or after some time when enough statistical information about ad performance is gathered by Google. The latter takes some time and it would surprise me for new sites that one or two weeks is enough to make this decision.

The strange thing is that many people see EPC falling after some days or a week, too late for an algorithmical calculated smart pricing, and too early for a statistical one. The effect could in my opinion be caused by human intervention of advertisers checking the referring publisher sites and withdrawing their high paying ads if they don't find the sites appropriate.

A related question: Does someone remember how smart pricing affected sites when it was just introduced? I wasn't in the AdSense/Words business at that time so I would like to hear some comments from the early users. Was the EPC falling from one day to the other, or was there a gradual lowering of EPC in a period of days or weeks? That could give some evidence if smart pricing is algorithmically or statistically determined.

I can sure see how publishers might blame the results on Smart Pricing.

It is not strange that they blame Google instead of the advertiser. In the beginning the advertiser had only two options. Opt-in, or opt-out of the content network. After many requests from advertisers and advertisers leaving the content network because of lack of control, AdWords added the possibility to filter 25 sites per campaign. A beginning, but still not enough to really influence the ad inventory for the publishes. Now they changed the number of sites in the filter to 500, advertisers can have influence on where their ads are displayed and this can have a huge influence on the ad-landscape in some publishers' niches. Also don't forget the option EFV mentioned in another thread that advertisers can now bid with different maximum bids in the content network. Many AdWords users still don't know or don't use these features to their full extent. Once they will be used more, the content network will change and some publishers will certainly suffer more than others.

[Chapman]

lammert-

One of the most insightful and educational posts I've read regarding the interaction of the two business perspectives.

Many thanks!

[europeforvisitors]

A related question: Does someone remember how smart pricing affected sites when it was just introduced? I wasn't in the AdSense/Words business at that time so I would like to hear some comments from the early users. Was the EPC falling from one day to the other, or was there a gradual lowering of EPC in a period of days or weeks? That could give some evidence if smart pricing is algorithmically or statistically determined.

I saw a drop almost overnight when smart pricing was announced at the beginning of April, 2004. My eCPM for April, 2004 was a little over 77% of my eCPM for March. The eCPM gradually crept back up, probably because of increased demand for AdSense ads as a result of smart pricing.

[lammert]

A drop overnight is not the same what people now experience. It is however not a proof that the current gradual decrease of earmings is not caused by smart pricing. When smart pricing was introduced, many statistical information of the AdSense sites and pages was already known, which may have simplified a overnight implementation of the pricing system. With new sites that statistical information has to be gathered first.

[21_blue]

I've not noticed this until reading this thread, but some of my highest EPC channels are those with the fewest words. I guess the more I write, the more likely I use a negative keyword that depresses my income!

[annej]

I too had a big plunge in income when smart pricing was introduced. Very gradually my earnings have come back and surpassed what I was originally earning. But that included a good deal more content building during that period.

Recently my earnings have dropped again though not as severely as when smartpricing was introduced. I don't have time to go back to my stats but wasn't it the summer of 2004?

[europeforvisitors]

Smart pricing was announced on April 1, 2004.