Domain Search: Adapting to New TLDs, Part 2

In my last post, I explored the changes that need to be made to domain name searching in order to take advantage of new TLDs. At Rightside, we consider relevancy to be the key to a successful domain search experience that benefits both customer and registrar. To that end, we've spent the past 2+ years developing a relevancy-based domain search solution called the Domain Suggestion Service, or DSS.

DSS was designed specifically with new TLDs in mind. As soon as new domains started to become available, we realized that it would be essential for the industry to figure out how to return relevant search results to customers. As the owner of both a registry and registrar, we couldn't afford to sit back and hope that someone else developed a solution. We knew that we'd have to create our own - so we assembled a team with experience in the domain industry, search, and data analysis, and got to work. Two years later, we have a product that we're very proud of. After extensive testing and development, we have begun to offer DSS to other companies, with Dreamhost and Dynadot already using the service, and several other exciting partnerships in the works.

So, what is DSS?

DSS is a simple but highly customizable API that delivers lightning-fast relevancy-based search results. But how does it actually work?

As a registrar, when someone searches for a domain on your site, you typically only get one or two words to go on, three if you're really lucky. In many cases, you don't really know anything else about the customer except what they search for. So how do you take two words and figure out the best results to show? You start with the TLDs. There are over 600 new TLDs currently available to customers. One of the ways that we identify the most relevant TLDs for a search is our library of over 100,000 editorially curated keywords, which are linked to every new TLD available. If a search contains one of those keywords, we know to serve up the associated TLD as one of the results. Even with 100,000 keywords, we know we can't cover everything, so we supplement our keywords with data from a variety of sources, finding additional links between TLDs and the search terms.

In addition to the clues we can gather from the search term itself, we use the data we can know about a customer. If a customer is searching from a location associated with a geographic new TLD (such as .BERLIN or .NYC), or ccTLD (like .CA or .FR), we can include that TLD in their results. Often, customers are already logged into their accounts when they perform a search, so we are also working on a new feature that incorporates data about previous searches, purchases, and any other information they might have entered, to provide even more personalized search results.

We also leverage all of this data to determine the best way to categorize search results for users. This allows us to return a value noting whether the search fits into a specific vertical, such as the legal industry or a personal name search. Name.com uses this feature to redirect customers to a customized 'personal identity' experience that is tailored specifically to someone searching for their name.

With all of the work we have put into discerning what a customer is looking for, one of DSS's strongest features came out of the realization that we need to know when we can't know something. Sometimes a search just doesn't have any clear meaning. We found that in some cases, a pure relevancy-based approach was not the optimal solution. When the system had nothing to go on, it resulted in a poor user experience. We figured out a way to identify those types of searches (which we termed 'generic') and treat them differently from searches with a clear meaning ('vertical'). Now, instead of getting a haphazard list of TLDs that aren't strongly correlated to the search terms, generic searches get a tailored list of broad-purpose new TLDs that cover a range of uses.

Even with generic searches, we still allow several opportunities for relevancy to influence the results. Take the search 'brokenankle,' for example. We don't have a lot of information linking that search to any TLDs, so it is classified as a generic search and receives a set of broad-purpose TLDs in the suggestions. Separately, the words 'broken' and 'ankle' don't give us a lot to go on - but combined, our frequency data is able to link the terms to several medical TLDs which get mixed in with the broad-purpose TLDs to provide a well-rounded mix of suggestions for the customer.

These are just a few of the core features that make DSS a powerful solution for providing best-in-class relevancy-based domain suggestions. There are many more features to highlight, and we are adding more all the time. DSS is now available to the public, so if you are interested in implementing it on your site or learning more about its capabilities, please reach out to us at [e-mail].