Google Adsense penalizing popular sites?
March 5, 2007 Posted by Bryan in : News , trackback,
Email this post
CENTRAL TEXAS — Data on several sites maintained by the operators of Texxors.com suggests that Google may be manipulating the profits webmasters receive from Google’s Adsense program. Our analysis shows statistical evidence that as a website becomes more popular, it receives less earnings per click (EPC) from Google Adsense. Since the EPC is determined by Google’s algorithms prior to the ad being served to a page, this suggests that Google may be intentionally or unintentionally manipulating EPC to increase their profit and/or Adsense participation. The method appears to be similar to the “throttling†practices that landed online movie retailer Netflix in legal hot water last year.
In our data, early earnings per click were on average 64% higher than those occurring later, after the sites had experienced growth. This drop is equivalent to a decrease of around 36 dollars per every hundred earned.
(continues below)
Google Adsense is used by individuals and large companies alike as a source of revenue. Several of the top 100 Internet sites in the world have Adsense ads on their pages. Clearly, the practices and design of the Adsense program affect the income of websites of all sizes.
If Google is cooking the books with its implementation of the Adsense program, it has the potential to influence revenues across the Internet. Indeed, our data shows that there are statistically verifiable differences in the average EPC that a small, low traffic site will likely receive compared to the average EPC a more popular site will receive - in Google’s favor.
I’ll begin by discussing the general trend that as site traffic increases, the average EPC of the Adsense ads served to it decreases. (Note: I’ll try to keep the language here simple and jargon-free for those who may not be familiar with statistics. For those that are familiar with statistics and may have concerns about the lack of information regarding standard deviations, residuals and the like, you can email us for that information.)
The graph below shows the average earnings per click for the six sites that we at Texxors.com manage (texxors.com, uncomfortablesituations.com, workchest.com, datechest.com, datechest.blogspot.com, workchest.blogspot.com, uncomfortablesituations.blogspot.com). As Datechest.com and the blogspot sites are new, most of this data was taken from activity over the last 19 months on texxors.com, uncomfortablesituations.com, and workchest.com, which came online in Aug ‘05, July ‘06, and Feb ‘07 respectively.
To the untrained eye the trend may be difficult to distinguish. But if you hide the two large peaks in the graph (place a finger or piece of paper over them), it is easy to see that the general tendency is that as clicks per day increase there is a corresponding decrease in the average earnings per click.
The orange horizontal line indicates the 20 cent per click mark. You can see that prior to the 10 clicks per day threshold, the average EPC is equal to or over 20 cents/click at 3 points, but after the ten click per day threshold it does not reach this point again. It turns out that if we compare the average EPC on days when the sites received 10 or fewer clicks to days when the sites received 11 or more, there is a statistically verifiable difference (Stats people should see note 1).
When the sites received a total of 10 or fewer clicks per day, the average EPC is 18 cents per click. Above this number of clicks the average EPC is 11 cents per click. That’s a decrease of 40%. In other words, the average EPC on days with 10 or fewer clicks is 64% higher than days with higher numbers of clicks.
There is a similar trend with page impressions. The table below presents a graph of average EPC by daily page impressions. This graph is even noisier, but looking at the farthest left section, representing the lowest number of daily impressions, it is apparent that there is an initial sharp decrease in average EPC before it begins to level out around the 30 impression mark.
To make this clearer, the graph below presents just that far left section of the previous graph, making the downward trend more easily discernible. (Stats people see note 2).
Finally, as anecdotal evidence, the table below presents today’s information (3/3/07) taken directly from the Adsense reports for one of our new, growing sites that still has low enough traffic to see the difference.
| Adsense traffic for Datechest.com - 3/3/07 | ||||
| Ad Block | Impressions | Clicks | Earnings | EPC |
| Top banner | 305 | 19 | $2.59 | $0.14 |
| Profile detail | 40 | 4 | $0.88 | $0.22 |
Notice that the average EPC for the ad block with higher impressions and clicks (”Top banner”) is 14 cents per click while the average EPC for the ad block with lower impressions (”Profile detail”) is nearly 22 cents per click. Both of these ad blocks reside on the same site (and sometimes the same page) and should be serving similar ads.
Clearly, whether you judge a site’s popularity by clickthroughs per day or page impressions, sites (and individual web pages within that site) that are more popular appear to be making less money per click from Adsense.
The inevitable question is “Why?” For that I don’t have an answer except to say that it makes good business sense for Google. The Adsense program was once essentially the only program of its kind, and Google was able to bask in its monopoly. But now other big name Internet companies are making a push to offer similar services. One particular competitor of note is Yahoo!’s Content Match® program. Although it is still only in limited release, its emergence signals that Google’s Adsense can no longer get by with the “If you don’t like us, where else are you going to go?†approach, but that it will actually have to compete for its users. It therefore makes sense that Google might do whatever it can to hook new users on their services first, with the idea that once the average user has learned one system and implemented it into their site design, they won’t want to switch to a competitor over a drop of a few pennies per click.
And of course, to stay profitable and competitive on the other side of the coin (the PPC marketing game), Google has to be able to offer low bid rates to advertisers. It could be that like many big-box retail stores, Google is willing to take an initial hit in one area to continue profitability in another (a la many electronics retailers). By serving higher EPC ads to lower traffic pages, Google hooks new users and provides more outlets for their Adwords PPC advertising.
In fact, Google could hold a fairly defensible position that they have nothing to do with the drops in EPC by using rhetoric about seasonal bid fluctuations and the market based nature of the PPC advertising which partially drives Adsense revenue (It took a lawsuit to get Netflix to admit to throttling unprofitable customers). The first couple of times I noticed this trend I attributed it to these facts myself. But as we’ve added sites at different times of the year with focuses on different markets but seen similar trends in Adsense EPC, it’s become clear that this wasn’t the case. And after looking at data from these sites using tried and true statistics, I feel pretty confident that something is going on. Whether it is intentional or not, and whether or not Google is attempting to keep it a secret are the big questions and the big issues.
It is impossible to answer these questions with the limited data we have. In fact, without having a better idea of how this might be happening, it’s difficult to even analyze the data at the correct level. Given that this analysis was exploratory, my approach was to consider aggregate data from all of our sites. This design has flaws when making assumptions about the EPC of individual ad blocks being displayed on each page. In fact, it’s even unclear from this data what really drives the decrease. Since page views and clicks are related it could be an increase in either one (or both) that triggers the decrease in EPC.
What this data can show is a statistically significant difference between the average EPC above and below the thresholds of 10 clicks per day. Additionally, it shows a negative correlational relationship between average EPC and the number of page impressions per day until the average EPC bottoms out at around 30 page impressions a day. But as any statistician will tell you, correlation does not prove causality. The cigarette industry rode that train for decades as a means of fighting assertions that cigarettes caused cancer.
Regardless, this information has some important implications for web designers—one of the most important being that a few clicks on low traffic pages are likely to be more profitable than many clicks on high traffic pages. Based on this data, it appears you could make more from 3 pages that got 5 clicks each a day than one page that got 20 clicks per day.
Depending on how Google is covering that initial increase in cost it may also have implications for PPC advertisers. Granted, a clickthrough is a clickthrough, but when you purchased that click did you imagine you would be paying more for an ad on a low traffic webpage than on a high traffic one? (Interestingly, while this article was in progress a NY Times article spawned a debate over whether the average person’s blog was a better place to advertise than larger sites. For more information read this article at Seth Goodin’s blog.) If Google is serving higher bidded ads to smaller sites then some advertisers will certainly be unhappy with that and the forthcoming changes to Adwords based on Quigo’s model cannot come to soon. If Google is taking that initial hit by increasing the EPC from its own pocket then that is different, but we really don’t know.
Unfortunately, I really have no call to action to close this piece. Until Yahoo’s Content Match goes into full release there still remains no equivalent alternative to Adsense. I will personally continue to use Adsense, specifically for the reasons listed above – because I don’t want to hassle with changing all of my sites for a few pennies. But as I add a new site and see the huge returns for the first few clicks only to watch them drop as the site gets more popular, I am reminded that for every 100 clicks I receive on a medium traffic page I am making about 20 dollars and losing about 7 compared to what I would be making if those clicks were distributed over lower traffic pages – or if there was not such a significant drop in earnings per click.
Note 1: Based on Independent Samples T-test assuming unequal variances. N=149, p<.05
Note 2: There is a significant curvilinear relationship at the p<.05 level. N=25, R^2=.28, F=4.23, const.=.865, b1=-.07, b2=.002.
Author’s Bio: Bryan holds an advanced degree in social research and currently teaches research methods at a small liberal arts college in the south. He has consulted for the U.S. Institute of Peace and worked as a researcher with various state and educational institutions. His previous written work has been published in both the academic and popular press. Despite his wife’s urging he still refuses to change his middle name to “Kick-ass researcher.”
Distribution: You may re-publish this article in print or online, in its original form, without further permission. When doing so, please provide a link to the original article (preferred method for online) or texxors.com (preferred method for print).
Last 3 posts by Bryan
- 3 things Cuil got right
- How to increase your electric bill and gain social status
- Samsung BlackJack Internet Sharing with Windows Vista
Comments»
Google Adsense may be penalizing high traffic sites…
Google Adsense may be either penalizing high traffic sites or using bait and switch techniques to hook new users according to this analysis of Adsense revenue for six sites….
Interesting analysis. But as you write, it’s hard to make any definitive claims based on data from just a few sites operated by the same entity. It doesn’t help that Google doesn’t talk about its algorithms either. Maybe Google actually adjusts its pricing based on the quality of the traffic, and perhaps your sites have higher quality traffic initially.
Either way I don’t think Google is ‘a cheap bastard’ as the Reddit headline for this article says. They get to set their own prices and you get to choose whether you want to do business with them or not.
Alexander, thanks for your comment. Certainly we’re not trying to make any definitive claims, but maybe (if anything) to get people to consider that this may be happening and to make informed decisions as a result or to at least look into their own data. That, of course, is the purpose of exploratory research - to take that initial (albeit sometimes slightly misplaced) step.
The reasoning to me is less important than the fact that it might be occurring. For some people choosing a method of bringing in income to their site is an important decision when planning their site. If they are testing various methods in the early stages and they make a decision based on information that is not truly representative of what they can expect on average, and that information is not truly representative of the norm because of intentional manipulation, then “cheap bastard” or not, that’s not a very honest business practice. To me it is akin, as it has been put, to bait and switch practices that can get you in serious trouble as a “traditional” retailer.
Let me be clear, I/we have nothing against Google. What they have done to revolutionize the internet is amazing and we have the utmost respect for them. But that doesn’t mean that now that they have to worry about stock holders and the bottom line, they will be immune to the negative aspects of human and corporate nature. Hopefully they’ll remain as good and as “little person” centered as ever, but sometimes we have to be willing to critique the Netflix’s, TiVo’s and Google’s out there to keep them from turning into the AOL’s, Microsoft’s, and Time Warners’ out there.
[…] Watching my Adsense traffic from over two hundreds of sites I’ve noticed lately an unusual drop in performance. My average adsense income per site had dropped almost in half. The truth of the matter that I honestly considered this to be a natural phenomenae accuring as a result of lowering traffic and never bothered to look into my server logs. Not until I stumbled upon this article which thanks to it’s authos is published under very liberal licensing so I will draw upon it rather heavily since there is no need to rephrase something with wich I absolutely agree. Here is the punchline: Data on several sites maintained by the operators of Texxors.com suggests that Google may be manipulating the profits webmasters receive from Google’s Adsense program. Our analysis shows statistical evidence that as a website becomes more popular, it receives less earnings per click (EPC) from Google Adsense. Since the EPC is determined by Google’s algorithms prior to the ad being served to a page, this suggests that Google may be intentionally or unintentionally manipulating EPC to increase their profit and/or Adsense participation. The method appears to be similar to the “throttling†practices that landed online movie retailer Netflix in legal hot water last year. […]
Hello! Good Site! Thanks you! gkuqobngrrwim