Wednesday, December 19, 2012

Ranking domains

Ranking plays a big part in search engine features to produce relevant results. While you may have several forms of ranks such as page rank, static entity rank etc., lets focus on one of these - the domain ranks. Is it possible to rank all the domains in the world? What will be rank them upon - popularity? viewership? political or geographical significance? Let's analyze it a little bit here.

Search engines care to provide very relevant results. Assuming that it has understood the query correctly (i.e., the user intent behind the query), it will have thousands of relevant pages as candidates for a result. To sort this result, one of the parameters will be the reputation of the domain.

There are several details known about a domain such as number of in-links (links from other sites to this domain), out-links (links from here to other domains), number of unique visitors per day, number of ip addresses behind the domain, page load time and the volume of content indexed from the domain. If you are a domain holder, then its worth you worry about all these properties to show up right at the top. Some domains are very friendly to search engines by having pages of very high html standards (including meta tags, sitemaps, etc).

Challenges include differentiating domains to sub-domains. For example, every sub-domain of blogspot is equivalent to a domain. Also, there are link exchange websites that will put a link to your website from theirs. There are advertisements that show up everywhere. These domains make it difficult to meaningfully extract a domain rank from the characteristics of a domain.

Even if we group domains into a bucket of say 10, we will be doing a useful job. We do not need every domain to carry a distinct rank.

Remember that the most searched and clicked through site need not be the largest site. Neither may it have too many in-links. Some domains may be geography dependent. For example, shows in your local theatre is probably most searched in your locality and should be ranked very high if the query and intent belong to that locality. Typically a convergence of query, intent, geography and static-reputation converge together to give a great search experience.

Saturday, December 8, 2012

About this blog

I will have ideally liked to name this as "Science of Search". However, I found a beautiful blog by that name already existing (scienceofsearch.blogspot.in) and hence had to settle for artofsearch. Anyways, Search Engine is a beautiful human invention that has loads of art and science behind it. In a series of posts, I will try to bring out my learnings on this space. Wish me good luck!