Internal Search Thoughts

I have a passion for internal search. I’m talking about the internal search you find on websites so you can search for the products or information that you are looking for. My last employer hired me because I was the only candidate that had a passion to make internal search better, and not just from an SEO standpoint, but from a usability perspective. I have found that internal search just isn’t as cracked up as it should be and doesn’t work the way a visitor wants it to. Here are some of my thoughts to make it better.

There are two facets that I want to cover in this post: Internal search from an analytics angle and internal search from a programmatic angle.

Internal Search From an Analytics Angle

Analytics can give us a lot of information about user behaviour. We can see how people use a website — when and where they leave, how long they stay on a page, what the most prevalent exit pages are, the most visited pages from a search query, and more. So, here is my question: Why aren’t we using these analytics to provide a better search for users?¬†Why can’t we make internal search more artificially intelligent by using vistor behaviour?

I think that using analytics to help drive internal search results will help drive the search engine to become more intelligent and thereby creating a better environment for the user. Why? Because they know what they are looking for, not us.

So, here are some of the ideas that I have come up with to make internal search better from an analytics angle by applying user behaviour.

  • Most Visited Pages From a Search Query – I have found this to be most useful when search results are presented as the users will click on the most relevant information. By applying this type of user behaviour in the algorithm, the most relevant information will soon rise to the top of the results. I have found that this is most evident with the consumer medical site that I worked on. The users were looking for the most relevant information, but the content that was presented at the top was not the most relevant to the users’ queries. The users would click on on results halfway through the page, and then move onto the next page to find the information they are searching for. To me, this presented an issue with the internal search as it was not providing suitable information to users. However, by applying the analytics from the user behaviour into the search algorithm, this would allow the best possible results to filter to the top.
  • Most Exited Pages From a Search Query – Why are users leaving the site after searching for specific terms? What terms are creating the most significant exit pages? When a visitor comes to a site, they are looking for something. When they use the search box, they inherently feel that they will expect the the results page to return the information that they have been searching for. If those result pages are not exhibiting the results, the users will leave for a better user experience. In this, we can apply a more deeply faceted search to keep users on the site so that they can find what they are looking for. By analyzing the search terms that make users exit the site, we can investigate the terms to create better content for the search terms.
  • Time On Page From a Search Query – The metric determines how long someone is on a SERP (search engine results page) and/or on a product/article page. Are they bouncing off the product/article page and looking for something else, or are they consuming the content and moving through the buy cycle? Are they consuming the content that is presented or are they bouncing off and trying to narrow their search? By asking these questions, we can add this as an indicator that those results are not right and need to be moved down while the ones that have indicate more time is spent on the page should be moved up.
  • Search Query Exit Pages – This metric shows that the content that was consumed was not what the visitor wanted and decided to exit the site altogether. If that’s the case, then that content needs to be moved down as it’s an indicator that the content is not relevant to the search.

These are just a few of the ideas that I have come up with so far about implementing user behaviour into the search engine to make it more artificially intelligent.

The next portion of this post will be from a programmatic angle. Read my Internal Search Thoughts Part 2.




No comments so far.

Leave a Reply
  (will not be published)