of Gator Advertisements and Targeting
Benjamin Edelman - Berkman Center for Internet & Society - Harvard Law School
[ Background - Motivation - Methodology - Research & Analysis - Testing - Disclosures ]
research project is complete, and the testing system now provides only archived
data. Details in Disclosures section.
The Gator Corporation designs software to display advertisements on users' computer screens, triggered in part by the specific web sites users visit. The author has developed an automated method of determining which specific advertisements Gator has associated with which web sites, data that may be helpful to web site operators, policy-makers, and others in assessing Gator's practices. This article offers listings of more than eight thousand specific sites targeted by Gator as well as analysis of the advertisements shown. An interface is also available to let interested Internet users to test Gator's advertisements on their own.
The Gator Corporation, a California company founded in 1998, calls itself "the leader in online marketing" and reports on its web site that it has more than 30 million users. As Gator explains in its "Quick Tour" (page since replaced by Gator; preserved by archive.org; preserved by author), Gator's software is installed on users' Windows computers. This allows Gator to show advertisements while users are viewing any site on the web, and Gator promises advertisers (page since removed by Gator; preserved by archive.org; preserved by author) that it can even display their ads when users visit competitors' sites. Gator's ad formats include pop-ups, pop-unders, and sliders.
According to Gator, its method of displaying advertisements is more effective than others'. Gator says it offers advertisers a clickthrough rate of more than 10%, purportedly more than 35 times higher than average web sites. Gator attributes this success to "deliver[ing] special offers precisely related to the users' interest at the time they are making purchasing decisions or shopping for relevant goods and services."
Gator's "Rates" page indicates that its campaigns begin at $25,000 for a promotion of unspecified duration. A Business 2.0 article suggests that Gator's rates are significantly higher than the ordinary cost of placing an advertisement on a web site by contractual agreement with the operator of that web site.
Gator's activities have prompted a number of legal challenges. Facing complaints from the Internet Advertising Bureau as to Gator's activities, then including banner ads that tended to cover web site operators' own banner ads, Gator in 2001 sued the IAB and subsequently settled. In 2002, Gator was sued by media companies including the New York Times and Washington Post in a case in which the author worked as an expert. (See declarations and other case documents as well as related disclosures.) Gator has also been sued by LL Bean, Lending Tree, Six Continents Hotels, and UPS, among others. Details are in the pending suits section of the author's related "Adware": Research, Testing, and Suits index.
Web site operators often wish to learn whether their sites are targeted by Gator advertisements. In the past, answering this question required installing the Gator client on a computer and checking whether Gator displayed ads subsequent to repeated requests for the specified web sites. But this approach suffers from at least three distinct deficiencies:
With these deficiencies in mind, the author set out to devise a more efficient way of determining whether Gator had targeted its ads at a given web site and, if so, what ads were targeted at what portions of the site.
The author posited that the Gator client likely communicates frequently with Gator servers in order to report to the server which web sites a user is viewing and in order to learn from the server which advertisements to display in response. To determine the specific means of communication between the Gator client and Gator servers, the author installed a packet sniffer to monitor all communications between the Gator client and Gator servers, as shown in the diagram to the right. (According to Telecom Glossary, a packet sniffer is software that enabled "monitoring network traffic in order to recognize and decode certain packets [data transfers] of interest.")
The author observed that the Gator client sends standard HTTP requests (both GETs and POSTs) to Gator servers and receives standard HTTP responses in return.
The HTTP responses received from Gator servers fall into at least three distinct categories:
UPDATE: As of May 30, Gator's "group" file responses no longer contain group names or descriptions Instead, the group files contain only gibberish txt and a copyright notice. The testing system has been modified to remove this information.
These methods of data collection do not speak to the frequency with which any given advertisement is shown on a targeted web site. However, empirical testing confirms that most ads found via these methods are in fact shown frequently -- more often than not, by and large -- on computers with Gator installed that visit the corresponding portion of the targeted sites and that have not recently visited related sites that trigger the same or similar advertisements. (The latter restriction is as described in the text of the first example rule in category 3, immediately above.) Experience therefore suggests that when a Gator response in category 1 (above) references an advertisement, that advertisement is likely to be shown to a Gator user who subsequently visits the targeted portion of the site at issue. Future research will seek to determine the method of prioritization among competing advertisements, the required times (if any) between advertisements, and the other specific conditions that trigger advertisements or that cause their suppression.
In the course of determining the specific methods of communication used by Gator, the author reviewed a large number of specific configuration files transferred via ordinary HTTP from Gator servers to the Gator client. However, all data interpreted from these files was found in plain text, without encryption of any kind.
Research & Analysis
This section provides a sampling of web sites found to be targeted by Gator, as well as analysis of the specific portions of these sites targeted and the specific advertisements shown.
The following links detail 8,348 specific web sites targeted by Gator for advertisement display:
Alphabetical listing (by second-level domain):
A B C D E F G H I J K L M N O P Q R S T U V W X Y Z numbers
Sites by top-level domain:
.ORG sites, .EDU sites, .GOV sites, .MIL sites, .INT sites
Gator seems to particularly intensively target certain categories of web sites:
Type of site / source
Sample targeted sites
% of sites targeted by at least
one Gator advertisement
Average number of Gator ads
targeted at each targeted site in category
Credit Cards Sites Source: Google search on 'credit cards' 65.5% 12.7 Targeted sites include americanexpress.com (43 ads), ftc.gov (2), bankrate.com (36), discovercard.com (2), hypermart.net (2) Movies Sites Source: Google search on 'movies' 64.5% 34.5 Targeted sites include imdb.com (14 ads), go.com (35), digitalcity.com (18), hollywood.com (17), yahoo.com (309) California Sites Source: Google search on 'California' 60.2% 8.8 Targeted sites include ca.gov (3 ads), berkeley.edu (5), usc.edu (2), caltech.edu (2), ucdavis.edu (4) Jobs Sites Source: Google search on 'jobs' 57.9% 12.2 Targeted sites include monster.com (19 ads), jobs.com (6), hotjobs.com (11), opm.gov (7), jobs.net (5) Refinancing Sites Source: Google search on 'refinancing' 56.5% 19.6 Targeted sites include bankrate.com (36 ads), cnn.com (27), householdauto.com (5), forbes.com (8), homestore.com (24)
Note that all analysis operates at the level of second-level domain name. URL listings were stripped to second-level domain names, then data was retrieved as to Gator targeting of each second-level domain name. In each category, the second column reports the proportion of sites found to be targeted by at least one Gator advertisement. The third column reports the average number of Gator advertisements targeted at those sites that are targeted by at least one Gator advertisement.
following link provides a partial listing of advertisers whose products or services
are promoted using Gator:
Testing indicates that, in addition to advertisements targeted at specific web sites, Gator also shows certain additional advertisements subsequent to user requests for numerous web sites. (Gator seems to prevent such advertisements from being displayed subsequent to user requests for certain sites, including Gator's own site and sites specified in court orders). As of May 2003, an additional 17 advertisements were placed in this category of general advertisements, including advertisements about replacement ink cartridges, cell phones, lotteries, business cards, and DVD rental. Using the testing system below, these ads can be viewed by specifying the string notrgs as the domain name to check.
Targeting to user profiles. The research described in this document entails replication of communication between Gator client and server using a fresh installation of Gator that has not been used to visit sites other than those necessary to conduct this testing. Whenever the Gator client checks with the Gator server as to advertisements to be shown (subsequent to user requests for a given site), the client sends the server a unique user ID and machine ID as well as the user's IP address, time zone, and zip code. Based on this data, and the associated user profiling that it makes possible, Gator servers could customize the advertisements to be shown to each user. Testing to date has shown no signs of profile-based targeting of this sort, and tests with different or invalid user IDs and machine IDs yield the same results as tests with the original IDs as set by Gator's installation software. Nonetheless, Gator's marketing materials prominently offer a targeting capability ("... based on past behavior"). This "past behavior" targeting may reflect profile-based targeting not yet detected by the author . Alternatively, it may refer only to display of an ad as a user leaves a targeted web site (or some number of minutes or clicks after a user leaves that site). Future research may attempt to more fully flesh out Gator's profiling systems by comparing the ads shown to users with divergent browsing histories.
This section allows interested users to test advertisement targeting by Gator. Simply enter the domain name to check into the specified field and press Submit. The subsequent page reports 1) the specific graphical advertisements and specific web page popups and popunders shown by Gator in response to requests for the specified domain name; 2) the specific host names, directory names, filenames, or URL keywords targeted by Gator for the display of particular targeted advertisements; 3) the specific category groupings in which Gator places that domain name.
This testing system provides up to five graphical advertisements for each requested site.
This testing system retrieves data directly from Gator configuration files, as understood by the author according to the research methodology described above. The author has performed extensive tests to confirm that his understanding of configuration files correctly maps to outcomes as experienced by ordinary users with Gator software installed. However, before taking action in reliance on these results, the author recommends conducting additional hands-on tests as well as manually examining Gator's configuration files as to the sites of interest.
All uses of this system are logged for future study, analysis, and publication. The results of selected requests may be merged with the author's prior data and reported via the links above.
Review some recent tests below:
Update as to status and availability of this testing system: As discussed in Disclosures, below, the author has accepted a consulting role in pending litgation against Gator. The author has therefore decided to reconfigure this testing system to provide only historical, archived data as to Gator targeting -- not current "real-time" results. For historical status information as to this site, including its blockage by Gator on May 22-23, see testing status page.
Update as to use of this testing system: In the first ten days of availability of this testing system, users tested more than 54,000 distinct domain names. Of these domains, more than 4,700 were found to be targeted for the display of at least one Gator advertisement.
The author's interest in Gator arose in part from a prior consulting engagement in which the author served as an expert to parties adverse to Gator in litigation. See Washingtonpost.Newsweek Interactive Company, LLC, et al. v. the Gator Corporation.
This report uses data collection methods that were developed at a time when the author was retained by no parties adverse to Gator. Since conducting this research, the author has been retained by parties adverse to Gator to perform work apart from the research reported and made available here. The author continues to receive additional inquiries from other parties potentially adverse to Gator. The author has also been retained by and received inquiries from parties potentially and actually adverse to WhenU, generally regarded as a competitor to Gator. Due to these consulting relationships, the author has decided to reconfigure this testing system to provide only historical, archived data as to Gator targeting -- not current "real-time" results.
This page is hosted on a server operated by the Berkman Center for Internet & Society at Harvard Law School, using space and resources made available to the author in his capacity as a Berkman Center student fellow for academic and other scholarly work. The work is his own, and the Berkman Center does not express a position on its contents.
The testing system uses thumbnails by ABCDrawHTML. The testing system previously used technical assistance from hellyeah! networks (questions to Emory Lundberg).
Last Updated: June 7, 2003 - Sign up for notification of major updates and related work.