Finding Similar Content on the Web (Guest Post by Or Offer)

By Or Offer, CEO of Similar Group (see bio below)

As the Web changes and gets bigger, finding relevant and interesting content becomes harder. Trying to solve that problem, Microsoft recently added a new function to IE8, a suggested sites feature. That is a great tool to discover more sites to visit, and along with Microsoft we can find other big players offering suggested similar site services. This post is dedicated to comparing and examining services available today.

Generally, the discovery niche on the Web has developed a lot over the past few years and recently became very popular (I believe that most of you will find it hard to remember that the first browser, Netscape, had an application of finding related content on the web).

In today’s similarity market we can find 5 strong players:

  • Google – Through “Related:” in their search engine
  • Microsoft – Through IE8 Suggested Sites feature
  • Alexa – Through their website or  toolbar
  • SimilarGroup – through their Search engine SimilarSites.com or their Firefox add-on “SimilarWeb”
  • Xmarks – through their site or through their add-on

Each one of these players uses its own different techniques and algorithms in order to supply the best results.

We have decided to explore the performances of each player by running a small test to see which one supplies the most relevant sites. To obtain accurate conclusions we tested sites in different languages, visitor traffic sizes, and themes. The results are surprising and will be detailed below.

Before getting to the test, I would like to explain a little bit about each player’s technology.

–          Microsoft offers a “suggested sites” tool, recommending new websites based on sites you have visited in the past. The application is located on their Favorites bar.

–          Google defines similarity between sites using factors including back links, popularity of the page that provides the links, etc.

–          SimilarGroup uses a unique algorithm that constantly analyzes the Web. Additionally, users can vote whether a site is similar or not to the original site.

–          Alexa suggests related sites according to different sites that where watched by same users. Assuming that ”If you liked X, you may also like Y”.

–          Xmarks offers similar sites according to its users’ bookmarks.

To clarify which one of these technologies provides the best results, we examined 8 sites: gigaom.com, youtube.com, ebay.com, mint.com, novinky.cz, wetteronline.de, ynet.co.il, and globes.co.il.

For the bigger sites, Google delivers many results, but the output is not very precise. Related sites to YouTube, for example, included Gmail and Wikipedia. However, it did manage well with long tail sites (low-traffic sites, usually in foreign languages), and gave pretty accurate results. A list of related sites to ynet.co.il, gave a satisfying list of newspaper sites in Israel.

Alexa doesn’t supply an actual list of similar sites, but suggests other sites you may also like (similar to stumbleupon’s method, but using a specific site instead of themes). It managed great with small, long tail sites, but didn’t give relevant sites for the bigger ones. Most of the pages related to eBay for example, were eBay sites in different languages (rather than different sites). YouTube’s suggested sites included Facebook and Google Mexico (which are also irrelevant).

Similar sites, a search engine created by SimilarGroup, managed to provide good results for big sites as well as long tail sites. The accuracy was pretty high for all the sites we examined. The amount of sites was quite small (10-20 sites, as compared to Google’s hundreds of sites per search), but thanks to the relevancy of the results, finding similar content was very easy. YouTube’s similar sites list included MetaCafe, Google videos and Vimeo (a site that most of the other lists lacked).

Microsoft suggests sites in a similar way to Alexa (people who visited X also liked Y), but does it more effectively. Although the feature suggests only a few sites similar to the current one, they are usually very relevant.  eBay suggested sites where Craigslist, Amazon and BestBuy (that is a very short list, but supplies the similarity that we looked for).

Xmarks had some problems with long tail sites and included many irrelevant sites, but managed great with big sites. Another advantage is the long list of sites (around 30 similar results for each site).

The following graph represents a divide among excellent, decent and poor from the entire results.

similarweb1

Here is the same graph by percentage divide:
similar web 2

In summary, Microsoft, SimilarGroup and Xmarks do a good job and deliver the information in a convenient way. If you are looking for a wide variety of sites, Xmarks is highly recommended. If you are looking for high accuracy in both big sites and long tail sites, you should definitely use SimilarSites, SimilarGroup search engine. And if you prefer to receive a short list of relevant sites, you should try Microsoft’s suggested sites.

***

Or Offer

Or Offer

Or Offer, age 26 with a BA in Business Administration, is the CEO and co-founder of SimilarGroup, a technology hive of applications powered by a proprietary similarity engine.  At 14, Or started a large and successful BBS, then 5 years later, built a network of profitable ecommerce sites.  SimilarGroup is his first start-up.

Eze Vidra
Follow me

Eze Vidra

Chief Innovation Officer at Antidote
Eze is the Chief Innovation Officer at Antidote, a startup helping patients search and match to clinical trial, to accelerate medical breakthroughs. Previously, Ezewas a General Partner at Google Ventures Europe. Before GV, Eze founded and led Campus London, Google's first physical hub for startups, and was the Head Google for Entrepreneurs in Europe. He's an experienced product manager and startup mentor. In 2012 Eze founded Techbikers, a non-for profit supporting children education in developing countries.
Eze Vidra
Follow me

Eze Vidra

Eze is the Chief Innovation Officer at Antidote, a startup helping patients search and match to clinical trial, to accelerate medical breakthroughs. Previously, Eze was a General Partner at Google Ventures Europe. Before GV, Eze founded and led Campus London, Google's first physical hub for startups, and was the Head Google for Entrepreneurs in Europe. He's an experienced product manager and startup mentor. In 2012 Eze founded Techbikers, a non-for profit supporting children education in developing countries.