A Study on PageRank Values of Paid-Result Search Engines
PPC Pagerank Values: Theoretical, Ethical and Business Issues.
Dr E. Garcia, Mi Islita.com
admin@mail.miislita.com
November 25, 2002
Last Edited: November 28, 2002
XHTML 1.1 Conversion: March 25, 2004
CSS Formatted: March 29, 2004
Topics
Available Now » Report 2
WPSS: A General Framework for Web Page Scoring Systems
Another independent study finds more flaws! Entire theoretical framework is questioned.
PROCEDURE
For this study, we did not include all PPC engines available across the entire Web but those recognized by destination sites such as the above two seed urls. We plan to produce a series of PageRank Report studies to include more search engines. We also plan to report PageRank studies in connection with sites offering topic-specific services in different market spaces. Research work along these lines is currently under way. The general procedure for such studies follow the one described in the next paragraphs.
To determine the PageRank value of the main page of a search engine, we simply entered the url address in the browser location (IE 5) as http://www.enginename.com (.net, .org, etc, when necessary) and recorded the PageRank value displayed by the Google Toolbar. By a main page we mean the page that is presented to a user when the corresponding search engine url is entered in the browser location. For the most part, the main page corresponded to the index page of the search engine; so the PageRank was recorded, accordingly.
In some few instances, the resultant document was a redirected page rather than the domain index page. This was the case of Overture, WebCrawler and AskJeeves. PageRanks of their index pages were not available and we have no choice but to record the PageRank value of the resultant page. In some cases, Google's Toolbar failed to report the PageRank value of a visited web page. We found that by simply closing the browser window, opening and using a new browser window resolved this minor problem, without further incidents. To verify our results, we performed the PageRank analyses by triplicate, with identical PageRank scores as previously recorded. We are not sure if the above anomally is a true bug in Google's Toolbar, the browser itself or the entire browser environment, but certainly is a client-side inconvenience. Users unaware of this anomally may mistakenly take it as a case of "Current page is not ranked by Google" . [However, in the particular case of some engines (e.g., Infospace), we were not able to obtain any PageRank value, even trying on different days. These engines were not listed in the Results section of our study.]
RESULTS
The results from this study are summarized in our two-table divide. Table 1 lists engines that follow Overture's PPC model with some minor variants. These engines don't seem to maintain a business relationship with Google. If a relationship with Google does exist, we were not able to confirm that online. Table 2 lists engines, regardless of their business relationship with Google or regardless of the nature of their services (free, paid or combined). Notice we have included Gimenei, a relatively new and less known search engine along with more traditional and well known search engines.
PageRank values of search engines and which at the time of the analysis maintained a direct or indirect business relationship with Google were labeled as Google Partners and identified with an asterisk in the corresponding table (Table 2). Search Engines that simply use Google's free internal site search services in their site were not considered "partners" in the business sense. Google has no revenue agreement for such free service, nor is there a market share issue involved when using the service. Google's free internal site search service seems to be a co-branding agreement rather than a revenue agreement and cannot be mistaken for the services Google provides to Yahoo, AOL, Infospace, AskJeeves and other portals.
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
ANALYSIS
A Note on New PPC engines
Being a new property on the web is very relevant to this discussion since new properties are more likely to have small link connectivity structures across the web, small databases and few authority sites pointing back to them. These are arguments in which PageRank framework is based.
Table 1 shows PageRank values of relatively new and old PPC search engines. Pay special atention to LookEngine, ICraveCash and ZineMarket. At the time of the study these PPC engines were listed by PayperClickSearchEngines.com as new search engines.
LookEngine.com, has a PageRank value of 3, equal to that of 7Search, which is not a new PPC engine. ICraveCash has a PageRank value of 4, equal to that of long-standing PPC destination sites SearchKing and Ah-ha. We placed SearchKing on Table 1 because it is wrongly viewed by several "authority" sites as a PPC search engine, even though they offer many diverse services.
ZineMarket.com has a PageRank of 5, higher than SearchKing and Ah-ha and trailing Overture by just one PageRank unit. We were not able to assimilate these findings based on "Web seniority". Certainly these findings seem to question link connectivity, database size and back link arguments.
A Special Note on SearchKing
Considering that PageRank is a metric which propagates link weights across the Web, a change in one PageRank unit in one destination site has a significative impact on the PageRank of other destination sites. Lowering the PageRank score by multiple units is, therefore, devastating for any site or prefabricated link structure on the Web. This effect is more notorious on structures gravitating around a central trunk (a seed link) and whose PageRank value has been lowered suddenly by several units. This is the case of SearchKing, which has filed a lawsuit against Google.
According with the lawsuit, "SearchKing's PageRank was 8 in July, 2002. In August of 2002, SearchKing began PR Ad Network, a branch of its operations in the business of placing advertising and links on highly ranked web sites, thus competing with Google. Once Google became aware of the fact SearchKing was competing with it and profiting from Google's page ranking system it purposefully devalued SearchKing, PR Ad Network and the web sites it hosts, within its source code of "formula". This resulted in SearchKing's PageRank dropping to four (4) out of ten (10)." (5).
In other words, within weeks of launching a marketing program involving the PageRank metric, SearchKing PageRank went from 8 to 4, even though during the same period of time their overall link connectivity structure across the Web (with SearchKing as the central "trunk") was preserved. This sudden change of 4 PageRank units seem to be accomplished with human intervention.
It has been reported by some SEO specialists that this drop in PageRank value was the result of previous "more than average" changes to how Google ranks and rates web sites. If so, it is very suspicious. First, is this Google's explanation of the facts or just rumors? Second, "more than average" doesn't help to clear the picture. Third, were "more than average" the number of sites penalized during the same period of time? Certainly, Google has the right to change its algorithmic rules. That is not even an issue. They simply don't have the right to target some sites they are displeased with.
The fact, (and reported by independent sites), that approximately during the same period of time, SearchKing's site was dropped from the first position in the list of search results for a query for "searchking" and only restored when SearchKing complained, suggests human intervention in Google's algorithmic technology. Can we remember when Microsoft was #1 in Google in connection with queries for "go to hell"? Someone may argue: "who will ever want to do a search for such queries?" That's besides the point. User's motivations when querying Google or any database is not even a valid argument. Google and Google users claim to use a technology which reports documents with the most relevant content. That's how Google has earned recognition in the industry. User motivations when querying a database is not even an issue for an algorithmic-based technology. Unfortunately, this subject is out of the scope of the present report and is not the central thesis of the study.
Why does Google seem to be so concerned about PR AD Network? SearchKing describes the service as follows:
"We're kind of like an online dating service. Our job is to help get sites with a high Page Rank that are wanting to sell advertising, together with marketers who see the value in having text link ads on pages that attract their target market AND could increase the Page Rank of their own site thereby helping generate even more traffic from other sources. We are matchmakers."
This does not sound like a "free-for-all" or "link-farm" structure. The process of matching and dating is about placing commonalities within the best possible scenarios. PR AD Network is an innovative marketing strategy that happens to use a metric an industry pretends to follow. In many ways PR AD Network attempts to validate a metric viewed in an industry as a "standard". Certainly the element of profit is in the picture, but that is a marketing issue, not a link connectivity issue.
Notice that a PageRank of 8 is the one assigned to many well known, well connected search engines (see Table 2). The above actions from Google can hardly be defended as based on "random walk" theoretical arguments or arguments based on link/weight propagation across the Web and certainly are contrary to the Scientific Method. SearchKing's lawsuit against Google does not seem to be a case of a small company trying to profit from the intellectual property of a large company.
It should be pointed out that search engine optimization companies (SEO) profit from search engine first-page ranking results and page rankings in general . 1st-page rankings is a metric derived from the technology of search engines, Google included. That is not any different from what SearchKing was trying to accomplish. They just launched a common sense marketing program that happens to involve the PageRank metric. Anyone with enough vision can now design any sort of marketing programs involving PageRank or even make derivatives of the "dating service" pioneered by SearchKing.
If Google's marketing department did not have enough vision to profit from PageRank or Google ranking technology in general, they cannot punish or go after others that did have that vision or will have the intention to do just that. If they did not want anyone to even think about profiting from the PageRank metric , why then make the Google Toolbar, PageRank and their metric results available to the entire Web community?
SearchKing's lawsuit, does not seem to be a case of "David vs Goliath", either. How about destination sites interlinked to SearchKing? Link weight punishment by association, (of innocent participants of link structures), may become a valid argument to be elucidated in the legal case of SearchKing against Google. In fact, this may become a career-defining case for any lawyer . It may become a case in which the entire PageRank framework, as a valid metric, "lands in the court bench".
Certainly, SearchKing affiliate sites and their fair-trade rights have been victimized by association, but these sites are innocent victims of a framework that claims to propagate and measure link weights in a fair fashion. Our impression is that after everything has been said and done, this case may be reduced to a matter of who will profit from PageRank, when, how and to what extent . That may be the "bottom line", too. As we can see, what works in a computer lab or R&D environment, not necessarily works in Wallstreet, "main street" or before a federal court judge.
A Note on Overture
Table 1 displays other interesting findings. Well known PPC engines such as Kanoodle and Overture received a PageRank of 6. The case of Overture, well connected on the web and a long-standing archrival of Google in the paid-results business is extremely interesting.
It can be observed that smaller, less known PPC search engines which in comparison with Overture take less market share away from Google have similar or higher PageRank values. This is surprising, considering that Overture has a major link structure presence on the Web, predates many of these engines and that many top search engines link back (give a "vote") to Overture but not to small, new, or obscure PPC engines. Overture's case seem to question the argument that link citation and link importance or link connectivity size is essential to achieve high PageRanks.
A Note on Non PPC Search Engines
Table 2 shows an interesting result in the case of Gimenei, a relatively new and less known search engine, that produces relatively small search results and which at least during this study was not in the PPC business. When we conducted our study, at times we could not get access to Gimenei. We don't know if they were "down" or not. With all, Gimenei's PageRank was higher than Ah-ha's and equal to that of Overture, Kanoodle and other well known PPC engines.
It can be noted that search engines not in the PPC business (Vivisimo, Wisenut, HotBot, MSN, Teoma, Lycos, AltaVista), which do not take PPC/paid-results market share away from Google, received PageRank values better than Overture or the rest of the PPC search engines.
A Note on Redirection
It can be argued that Overture received a PageRank of 6 because it redirects to another page. This does not seem to be the case. WebCrawler, (an Infospace property), redirects and has a PageRank of 7. AskJeeves redirects to Ask and has a PageRank of 8. Incidentally, AskJeeves has a strategic alliance with Google. According to Google, "Ask Jeeves and Google will share the more than $100 million in estimated revenue to be generated from Google's advertisers on Ask Jeeves' search sites over three years. Google's sponsored links service will complement Ask Jeeves' own suite of advertising products, including its proprietary paid listings product, branded ad units and paid inclusion products. Beginning in September 2002, Google's sponsored links will replace the third-party paid listings currently featured on Ask Jeeves search sites." (4)
Note More on the PageRank values of redirectional index pages can be found in the FEEDBACK FROM READERS section we have included with this report. As we mentioned at the begining of the paper and as labeled in our two-table divider, these are PageRank values of Main Pages. For the redirectional cases, these are for the page a user actually sees, not the actual index page. This has been stated all along in the Procedure section. Arguments presented in the reader's feedback section at the end of this paper reinforces the central thesis of our study; i.e., whether or not the PageRank metric is a valid or fair one.
A Note on Google Partners
Table 2 shows several interesting PageRank results in connection with properties owned by partners of Google. America Online LatinAmerica (AOLA), not a real "powerhouse" in the search engine world, (and a penny stock), has a PageRank of 8, equal to that of relative newcomers Wisenut and Vivisimo. It shoud be pointed out that AOLA is owned in part by AOL-TimeWarner, a Google partner. DogPile and MetaCrawler (both owned by Infospace, a Google partner) received a PageRank value of 8.
A Note on "Childs"
It can be noticed that AskJeeves and Terra, which respectively own Teoma and Lycos, both have lower PageRanks than their corresponding "child" properties, while AOL has a PageRank higher than its "child", AOLA. PageRanks are 9 and 8 , accordingly. AskJeeves and Teoma are Google partners. We cannot comment on how database size or ownership fit into this picture.
A Note on Portal Networks
It can be observed that WebCrawler and Starmedia have a PageRank of 7, even though Starmedia, (a penny stock), has a network of many directories. This disputes the argument that SearchKing has a low PageRank because it is a network of many directories; needless to say, Starmedia and WebCrawler are not considered "heavyweight contenders" in the search engine scene.
A Note on The Old Glory Days: WebCrawler and Excite
Surprisingly, Excite has a PageRank of 9, equal to that of MSN, AltaVista and Lycos. With regard to the PageRanks of WebCrawler and Excite, (Infospace properties), we need to stop and raise a valid question. A question that goes to the heart of PageRank as a valid metric. These two search engines were important factors in the "old" glory days of the Internet, but no longer are. The idea of remunerating old glory connectivity schemes in some cases but not in others makes no sense. WebCrawler PageRank value or its link structure across the Web, present or past cannot be as equally significant as that of Overture or other larger search engines or destination sites across the Web. In this sense PageRank is a static measure that fails to recognize link actualization across industries.
In terms of Web traffic from users that "never hit the back button", WebCrawler is not a major destination site today. At least, the PERCEPTION OF TRAFFIC is not there or can be measured as a significant "spike" to be compared with top destination sites, monthly reported by Nielsen//Net-Ratings, and others. Whether or not PageRank values of "stone age" search engines are meaningful is something to be argued, not in a computer lab but in an ebusiness environment. Certainly the results from this study seem to question the argument(s) that PageRank is all about state-of-the-art link quality citation or importance of link structures.
A Note on Yahoo and Google
Last but not least, only Yahoo, a Google partner, and Google itself are found with a perfect score of 10 PageRank units.
Update: A PageRank of 11 out of 10?
This update, added on 03/20/03.
According to an article written by Chris Raimondi, Google is assigning to itself a PageRank of 11 out of 10 in the Google Directory. Mr. Raimondi wrote "Did Google do this on purpose? It would appear so." The complete report is available at SearchNerd.com.
Sections
Available Now » Report 2
WPSS: A General Framework for Web Page Scoring Systems
Another independent study finds more flaws! Entire theoretical framework is questioned.
