Google’s new Knowledge Graph, also called “semantic search”, takes search beyond keywords optimization. In other words, it tries to understand the intent and meaning behind the search phrase and then provide a more relevant and complete picture to the user.
We recently conducted a few case studies on Semantic Search application with very promising short-term results.
Since the basis of Semantic Search technology is to provide relevant results, we reviewed and analyzed user sentiments across many review websites for particular hotels and their websites.
We then incorporated the top sentiments into the content of the website in a natural and very informative manner.
This was done for two websites: a hotel in New York and a hotel in Maui, both are full service hotels with great amenities that were fully described on the websites. However, after a deep analysis of the user sentiment, it was discovered that people were also talking about about several other features of the properties in great depth, including: the lounges, décor, outside ambience, large and inviting windows, etc
We then proceeded to weave the above findings within the content of the website – using similar key phrases that we found in the user sentiments. We also created new pages (where needed) to highlight more features and provide more information.
Short term results:
After two weeks of semantic search alignment, the New York hotel’s website revenue increased by 77% due to the further optimized, relevant pages. People coming to those pages
- Spent more time on the site,
- Viewed more pages,
- And significant decrease in bounce rate.
For the Maui hotel’s website, we saw a similar increase in metrics. The traffic, page views and revenue increased by over 50% in the short period of two weeks.
The key take away is Semantic Search definitely work! By making the content more relevant, we found that more people were coming to the website – driving enhanced website performance in search results. It’s evident that with the new Knowledge Graph, Google is attempting to correlate relationships between entities, just like an actual graph, and provide better results.
Contributed by Manisha Kumar, Director of Strategies