[ Home Page | Condensed Version | The Movie | The Book | News | Blogs | Feedback / Mail List ]

Search and guessability

I will testify that domain names play a minor and declining role in Internet searching, and that this fact refutes plaintiff's argument that Mishkoff's occupation of the <shopsatwillowbend.com> domain harms plaintiff's interests.

TThe domain name "guessing" process, whereby a user thinks of a company name and appends .com to it, is not common now. There are objective and documentable reasons for this change. What we might call "guessability" was a response to unique and temporary conditions when the World Wide Web (WWW) was new, between the years 1994 and 1997.

Note: The WWW protocol was developed at a Swiss physics research institute (CERN) in 1990-92. The first free, easy to use Web browsers only appeared in 1993; Netscape appeared in 1994.

At that time the WWW was overwhelmingly American and Internet content was overwhelmingly English. The Internet was much smaller. The number of registered domain names was 1% to 5% of the number registered now. Moreover, during the initial rush of WWW growth, domain name registrations were concentrated in the .com top level domain (TLD) to a degree that was excessive and historically unique. This concentration happened because .com was the only commercially oriented top-level domain that accepted registrations from anyone and everyone at a low price. Most of the other TLDs were restricted or much more expensive at that time.

Note: A more extensive historical documentation of this sequence is contained in my book Ruling the Root.

At the end of 1996, more than 75 percent of all domain name registrations in the world were under .com. That is, for any given name one cared to "guess," attaching .com to the end of it had a 3 in 4 chance of being the correct TLD (and a much better chance than that if you were searching for an American commercial company). Finally, the state of search engine technology and service was much less developed than it is now. Under these historically unique and irreproducible conditions, the search strategy of "guessing" a domain name was reasonably effective (although as far as I know there are no empirical scientific studies on how often users did this).

Current conditions render domain name guessing and domain name-based searching increasingly irrelevant. The sophistication of users, the search tools available, the number of domain names registered, and the expansion of top-level domains have all changed the situation.

  • Instead of the 1 million or so registered domain names in 1997, there are now over 40 million registered domain names. As the number grows, the viability of guessing declines.

  • Instead of 75% of all registered domains being concentrated in the .com top level domain, now less than half of them are. There are over 2 million registrations in the new generic top-level domains .info, .biz, and .us. These domains are growing more rapidly than .com. In fact, from October 2001 to August 2002 the .com TLD shrank from over 32 million registrations to 22 million. The new TLDs .info, .biz and .us, on the other hand, grew by about 5 percent each month.

    Note: Comprehensive statistics on the number of domain name registrations are available at the "State of the Domain" report, http://www.sotd.info/.

    Registrations in country codes also grew apace. Expansion of the number of generic top-level names and distribution of more registrations into the country code TLDs makes "guessing" less rewarding and hence less popular as a navigation strategy. A user has a lower and lower chance of knowing in which of the TLDs a given domain name he wants to find will be. As this happens, a domain name becomes more like a mnemonic telephone number - to use it, one needs to know it or be exposed to it, rather than guess it.

  • As the value of domain names as guessable search keys has declined, so has the number of speculative registrations and domain name disputes. My statistical study on domain name - trademark disputes under ICANN's Uniform Domain Name Dispute Resolution Policy showed that the vast majority of disputed .com names were registered at the height of the speculative domain name boom in the last quarter of 1999 and the first two quarters of 2000. Since then, the number of UDRP disputes has declined steadily, so that now the number of UDRP disputes per annum is half of what is was in year 2000. I have attached this study to the expert report.

  • Most important of all, the quality of search engines has improved markedly. Many users now report that search engines such as Google can reliably deliver the web site they want to the top of the search list based on a few keywords. One would, for example, never be able to guess the URL for my personal web page http://istweb.syr.edu/~mueller/. But if one types "Milton Mueller" into the Google search interface my web site will show up in the top three or four. Google did not exist five years ago.

  • Portals also play an important role in Internet navigation. Portal sites assemble links to commonly used, needed or desirable sites for an audience, and obviate much of the need for guessing domain names as a navigation strategy. For example, Yahoo!, AOL and Msn.com are portals. In a 2000 study of Internet use at Syracuse University, where we could precisely quantify traffic and domains referenced, we found that portal sites were by far the most commonly visited domains.

    Note: This study was proprietary research performed for Telcordia Technologies, hence the full report cannot be included in the record.

    Portals as a category led entertainment and pornography, no small achievement for a college audience. Other user studies indicate that the typical user confines almost 80 percent of their traffic to a small set of domains, to which they return again and again. Users store links to which they want to return in their "favorites" file (for Internet Explorer users) and return to them easily. This aspect of user behavior is rather inconsistent with the vision of a clueless user who scratches his or her head and painstakingly types in guessed domain names to navigate the Internet.

  • To my knowledge, there is no empirical research supporting the assertion that Internet users cannot distinguish between a domain name that points to a site they want and a domain name that refers to the same thing but is not the site they want. Indeed, arriving at the wrong web site because of typos, clicking on the wrong link, or some other mistake is a common experience of users. My impression (but I lack systematic empirical research on this topic) is that most users simply back up and keep searching, unless of course the site is deceptively confusing. To my knowledge, there is no research that refutes that common-sense expectation. The application to the Internet of the concept of "initial interest confusion" is primarily (if not exclusively) a legal construct at this point; it lacks an established corpus of social science research confirming its existence and defining its characteristics (such as how often users are "frustrated" and give up further searching as a result of finding the wrong site, or what counts as confusion). Internet research is a fairly new and dynamic area, so it is possible that relevant research exists somewhere and has escaped my notice, either because it is proprietary or because it is in an obscure journal. But I can say that there are no widely known, widely accepted empirical studies on initial interest confusion on the Internet.

  • A user study performed at Syracuse University under my direction surveyed a group of 52 students about their Internet search habits. They were asked what search methods they had ever used with success on the web. 100% reported using search engines successfully; 80% reported successfully using offline advertisements as the source of their URL, 76% reported successfully using hyperlinks or subject indexes on portals; 72% reported word of mouth or print listings.

    Note: Respondents were allowed to check more than one option.

    Only 68% reported ever having guessed a domain name successfully - making it 7th on a list of a dozen factors. Keep in mind that we were only asking them whether they had ever used that method successfully, even just once. When asked what methods were most likely to be unsuccessful, "Word of Mouth" and "Guessing" topped the list, with 56% of the respondents reporting a lack of success in the use of those methods for searching. Finally, when asked which was their preferred method for finding a web site for the first time, only 2% cited guessing a domain name. Search engines topped the list of preferred methods, being cited as best by 42% of the respondents. Hyperlinks was second. This was merely a pilot study done to test a research proposition for the development of a grant so the sample size is small, but it is one of the few sources of empirical evidence on this topic. That was in 1999; if anything, reliance on guessing has declined since.

  • In today's browsers, once you have correctly typed a domain name into the window in the past, the browser will store that information and automatically display domain names that match whatever you start to type into the URL window in the future. Thus, one no longer has to even finish typing in a domain - one can simply select it from a list in one's browser window. This obviously eliminates much of the guesswork in recalling domains. Thus, any user who guesses wrong and then goes on to the correct site will have the correct URL stored in their browser and easily accessible. Infrequently used URLs eventually get pushed off the list.

Next: Domain Names as "Source Identifiers"

[ Home Page | Condensed Version | The Movie | The Book | News | Blogs | Feedback / Mail List ]

©2002 Hank Mishkoff
All rights reserved.