I believe once everyone starts thinking clearly again after the last round of Penguin updates and focuses on their actual business rather than trying to find ways to game the system and mass produce activities that are not scalable we can get down to action that can actually move the needle.

One of the ways to find large god nuggets are in your site search queries.   While most keyword research best practices suggest you take a look at them I have found very few companies that actually do it.   For example, at a recent search conference I asked a room of 250 how many had even looked at their site search queries let along mine them and only four people raised their hands.   Talking to attendees after I round most don’t think it will yield any value, did not know where to get the data and others just did not think about it.

Why are site search queries important?

Site search queries are the most valuable search phrases you can have. They are the voice of intention of your customers when it comes to you. Site search queries represent the very products, services or information the searcher expects you to have since they are looking for it on your site.

The following are some ways you can use site search data to improve your search performance.

Mining Customer Questions

In my Big Data Mining session at SMX West I gave an example of a large tourist destination where we set up a process to mine their site search data looking for anything related to tickets.    In the table below are some of the key data points I will explain.

questions_keyword_matrix

The first thing we did was look for any phrases that were questions such as how, can I/we, where, when and what are all the basic questions as shown with their volume in the screen capture below.   We looked for variations of these with tickets, names of passes and other ticket related phrase.   What we found astounded us.  There were over 27,000 individual questions related to tickets.   We them took those questions and extracted the never of searches done using them which identified 600,00 searches in the current year.

We then wrote a script to test the site search appliance to see if any results were generated for each of the phrases.   We also did a check of Google for the top phrases to see if there was anything externally as well.    What we found was 60% of the queries were not generating any results.  Meaning if a searcher came to the site looking for answers to their questions on how to purchase, upgrade, exchange for an annual pass etc. nothing was presented.

monitizing_keywords

The next step was to review the questions to identify which were actually merchandiseable.   We looked for questions such as the following:

  • Where can I purchase family pass tickets online?
  • Can I upgrade my day ticket to an annual pass?
  • Can I upgrade my single park pass be upgraded to full access?

Once we finished our review, we found that 15% of the questions and over 225,000 searchers we should be able to monetize them.   The idea was simple, if they want to upgrade from a day pass to an annual pass they should be able to click a link and upgrade online or at least be given instructions on where to go at the park or who to call online.

The Marketing Director of the company looked at conversion rates and average sale for these types of products, day passes to annual passes, single park to multi park passes and assigned a average conversion value of $200 and a average conversion rate of 10%.

We then estimated that is we can convert those questions at that rate and value it would represent  $4.5 million dollars in revenue.  Now, we did assume that even with the horrible state of site search that many would figure out how to do these transactions elsewhere so we cannot count all the revenue but at least we had a number we can work with.

They created a number of pages of new content to address these questions and updated their eCommerce booking engine to accommodate upgrades and were are now able to measure the number of transactions that are made due to a positive result in site search for money questions.

Site Search Queries Improve Navigation

A few years ago a large B2B site updated their international home pages removing many of the product links and replacing them with interactive links based on where the visitor moused over.   We found that nearly 85% of the non-US users who came directly to the home page immediately went into site search as they did not know about this new design trend and did not interact with the page.

Based on the lack of clicks and the increased volume of site search, the UX team determined that the users were not recognizing or did not want to deal with the mouse over functions and were going into site search to find what they wanted.  They switched back to the original version and the site search rate dropped to less than 40%.   Little by little they adjusted the home page to include visual clues to the most common areas of the site which stuck the balance of creativity and function.

Identifying Non-Relevant Paid Search Queries

The second case I encountered recently related to paid search.  This B2B company was gating paid search visitors to only download a whitepaper.  The searchers clearly wanted something else with nearly 65% using site search to find what they wanted.   The PPC vendor wanted to remove site search from the navigation to stop this but it would only lead to increased use of back button.  We looked at the words and found the campaign was too generic and made adjustments to words and ad copy.  This did result in a decrease in clicks but a significant increase in engagement.   For those other words that were resulting in site search we increased performance in organic results and set up awareness campaigns to better interact with the searchers who wanted more specific information than getting the whitepaper.

This type of mining is simple in DataPrizm. You simply set phrase filters for each of the questions or product areas and then import all of your site search keywords into the database, the filters put them into their respective categories and you can use Cluster Analysis to see the different groups and take the appropriate actions.