BlogAds Analysis - Executive Summary
This past summer, just before leaving MIT for a teaching job outside Dubai, I collected some data for an empirical study on blog advertising. The specific objective of the study was to determine the strongest predictors of the revenue earned by ads appearing on blogs. In short, my findings are as follows:
(1) the number of weekly page views (WPV) is a much stronger predictor of weekly ad revenue and price than are either the number of inbound links or the number of blogs providing those links
(2) the number of ads has a negative impact on ad price and a largely positive effect on ad revenue, and
(3) the political orientation of the blog matters: on average left-of-center blogs significantly out-earn their right-of-center counterparts.
The study’s design is quite straight-forward. The first step was to gather data from the website of BlogAds, a North-Carolina-based company serving banner ads to over 1000 blogs covering a wide variety of topics and focus areas. One unusual feature of the BlogAds website is the wealth of detailed information it contains about each blog (page views, URL, topic/focus), its blogger (age, gender, state/country of residence, political orientation) and the blogs' ads (the number, kind, and position of ads, as well as their price). Two pieces of information- the price and the number of ads- were particularly valuable because together they can be used to calculate the total ad revenue for a given blog. I am aware of no study to date which has used this figure in analyses of blog or internet advertising.
With data in hand on 1357 ads appearing on 491 blogs, I next set about to develop and test a statistical model to estimate how well three groups of factors explain the differences in revenue across these blogs. Those factors were:
(1) the blog’s popularity, as measured by the number of weekly page views it receives and the number of sites that link to it
(2) the blog’s ad count, as measured by the number of ads appearing on the blog, and
(3) the blog’s political orientation, as indicated by my categorization of the blog as either right-of-center or left-of-center
(1) Traffic Is King. Page views are the strongest of the three predictors tested: it explains at least twice the variation in ad revenue that either the number of inbound linking sites, the number ads, or the political orientation do. Taken together, page views, the number of ads, and political orientation explain over 80% of the variation in the ad revenue, a very substantial amount. This indicates that the statistical model predicts ad revenue with a high degree of accuracy. Doubling the weekly page views results in a 50% increase in ad revenues.
(2) Space Matters, Even in Cyberspace. Holding page views constant, the average effect of adding one more ad is a 20% decrease in price and a 40% increase in revenue. Interestingly, the number of ads on most blogs is well below the number where returns to revenues become negative, i.e. the point where increasing the number of ads results in less, rather than more, revenue.
(3) Partisanship Pays, But Unequally. When controlling for both the number of weekly page views and the number of ads, politically-oriented blogs generate no higher ad revenue and command no higher ad prices than other blogs. However, revenues from ads on left-of-center (LOC) blogs are 40% higher than average while their right-of-center (ROC) counterparts’ revenues are 23% less. There are notable exceptions to this general trend, however. While LOC blogs dominate at both the lower (below 100,000) and the upper (above 1 million) ends of the weekly page views continuum, ROC blogs earn much more in the middle territory. Potential explanations for this difference, along with graphs, tables, statistical analyses, and a discussion of the other results, can be found in the full report.
UPDATE: Lately comes word that MSNBC is making a large BlogAds buy, over $1 Million. Check Outside the Beltway for more details.
Comments
It's my first visit to your website. After just a quick browse, I'm really impressed!
Posted by: Olga Ryrakhovsky | June 29, 2006 8:19 PM
Hasn't that always held true that people paying for ads want their ad to be seen by visitors. Unless you were planning on selling links for PageRank's sake, page views have always been important.
Posted by: Brandon Hopkins | March 29, 2006 2:10 AM
Jim
first let me say "Happy New Year"
Second, thanks for visiting my blog and for taking the time to leave a comment on the report.
Here are my thoughts about your remarks. Rather than consider the findings "biased" I prefer frame the matter as an issue of generalizability. Expressed as a question: "How generalizable are the results of the study?"
This is, of course, the question that is always raised about empirical studies that use data samples that include some proportion of a larger population. In order for us to generalize broadly about the results, it is important that we have an representative sample. I see your remarks as going directly to this point, i.e. how representative is my data sample. There are two ways to approach this question. The first is to ask what population am I interested. Since the study is about blog ads, not ads appearing on other internet sites, then there is a good argument to be made that the results apply only to blogs, bloggers, and blog readers and not to internet advertising as a whole.
The second issue, then, is whether I have a representative sample of blogs, bloggers, and blog readers. I did, after all, use data only from the BlogAds site and not from places like BriteAds or others that serve ads to blogs. Of course I like to think that the sample derived via BlogAds is representative, but this is something that is very hard to verify because the other blogs ad services don't share their data. that said, the fact that so many of the biggest blogs are included and that there is such a large number and wide range of blogs included makes me fairly sure that I have a good, i.e. representative, sample and that I can generalize about blogs, bloggers, and blog advertisements from that sample.
Your final point about readers self-selecting is a good one. To some degree everyone picks the websites they visit. In the case of blogs, I think this means that some people read some blogs and some read others. It sounds obvious but its important because it can mean that certain advertisers may be able to better to target their ads. after all, the most dedicated readers of the Daily Kos and Hugh Hewitt are probably NOT the same people. As such, there could be differences in either what kinds of ads appear on those blogs and, most importantly for my purposes, what price is charged for them.
Well, that's all for now. I'm off to see the new Harry Potter movie!
thanks again for visiting the blog and again I wish you a happy new year.
thoughtfully,
starling
Posted by: starling | January 1, 2006 4:35 PM
I think the study conclusions are interesting.
However, I wonder how much bias there is in the findings due to:
1) blog readers being a smaller pool than the general population
2) Readers self-selecting blogs they want to read.
Because as more and more people start to take an interest in blogs I suspect behaviour will change.
Posted by: Jim Symcox | December 31, 2005 3:42 AM
I find it very excellent how you have laid out this analysis. I would say the "hub and spoke" of your archetypal structures (as determined to be so, i.e. the limited influential bloggers with that of the high amount of traffic) is a correct assessment...
To control growth i.e. prohibiting counts of links as stabilizing and managing growth, will not prohibit growth. It will only separate the ranking system from the reading of it. Therefore the system that tried to prevent the escalation will in truth be inaccurate.
When you stifle growth you get competition between communities.. If I am reading the truth behind the current ranking system, it appears to hold concern of rank escalation by: "itself—to itself" (as with the "clique" of your archetypal structures)...
My assessment is that it is not for this reason, but the potential for blog’s $ worth which mediates this move...This explains the imbalance of "this blog with that" as appears a part of a business plan which did not lend proper analysis to the tides of blog growth escalation.
An analysis (at which you could provide) would be essential, immediately!
Posted by: MB | November 27, 2005 2:10 PM
Interesting study. Hugh Hewitt has long complained that Instapundit charges too little for his blogads; maybe he's holding down the right of center blogs revenues? I also note that the lefty bloggers are much more likely to have ads for political candidates and often run fundraising drives, which may create a feedback loop--advertising on Kos leads to fundraising by Kos, leads to more advertising on Kos.
Posted by: Brainster | November 16, 2005 10:56 PM
We have been tracking consumer acceptance of advertising in different media for nearly 20 years.
The latest: Blogs are trusted more than daily newspapers or network TV; consumers trust ads that appear in media they trust.
Americans' views about main-stream media Katrina coverage, for example, stated positively early in the crisis and faded over the next weeks. Blogs became the most trusted among major information sources.
Feel free to Email us for two Excel summary spreadsheets detailing the findings. Art@cnwmr.com
ArtemusS
Posted by: Art Spinella | November 14, 2005 11:32 PM