Hurol Inan: Web Analytics Consultant & Author

Analysis of the keywords that your customers use to search your website is one of the best insights you’ll get into your customers. If you’ve got some database skills, you can start analysing these keywords straight away - we’ll show you how.

To start with, two examples are provided below to give you a sense of the sort of discoveries that you should expect to make as a result of search keyword analysis.

On an Australian Federal Government website, we have discovered that one of the most frequently searched terms was not even in the Department’s portfolio, hence the website offered no relevant content. Many visitors, faced with unsuccessful search results, would send emails to the customer service department, who would have to deal with irrelevant enquiries. Our solution was simple - a new page informing the user that this Department does not deal with this matter, and providing a link to the correct department.

On a major professional services firm’s website, the analysis revealed that a great majority of people searched for office contact details. The fix was simple: include office phone numbers on the page footers.

Without the aid of a commercial product, I perform the search keyword analysis by uploading the search log files into a database (such as MS Access).

For analysis to be effective, collect search log files for at least a week, preferably for longer. Most commercial search engine products such as ISYS, Verity K2, Ultraseek and Autonomy produce log files in a similar fashion to web servers. They log the user’s IP address, timestamp of the search, search term used and the number of hits in search results.

After uploading the search log files into a table created in your choice of database with the same field definitions, you can analyse them in the following ways.

Analysis of top search terms
Using the query capabilities of your database, create a report that lists the top 50 terms along with the number of hits in their search results. Although it is advisable to go through each one of these terms by repeating the search by yourself to see how satisfied you are with the results, you should focus on very high and very low hit search terms first. Very low hit search terms may indicate shortage or even the lack of relevant content, while very high hit search terms may bring far too many search results than what the user desires.

Analysis of lowest hit search terms
To pick up the low hit search terms that were frequently used but missed by the top search terms analysis, query your database table by grouping search terms in order of occurrence and sorting it to identify the lowest 20 hit search terms out of a larger group of top search terms (say, 200). Again, go through these search terms individually – checking what your visitors are looking for and how well your website responds.

You will notice that many of the terms are OK, but some terms will need an improvement on their search results. Various problems may present themselves here. For example, an important page that should be displayed is not picked up by your search engine due to the lack of or infrequent of use of the search keyword on the page content.

Single keyword analysis
You will notice that many of terms used to search your website are composite words or phrases. With the reports discussed above, we looked at how well your search engine responded to exact terms. And these analyses may be skewed due to slight differences in terms. So it may be useful to decompose search terms into single words and analyse frequently occurring words. To decompose the search terms, I simply upload the log files into a temporary table with space as a field delimiter for the search terms – eliminating the need for writing a script program to do this.

Search context analysis
For more advanced analysis, you may want to understand the context of the searches through classification of search terms. This analysis would reveal the concentration of searches by relevant segments of the website and may assist you in determining where to focus your new content development. Considering the size of log files, it could be a major undertaking to map each search term to a context area. But remember, website analysis is not an exact science. A simple approach with a relatively high degree of accuracy would suffice in most instances.

Going back to the classification discussion, visually examine the search terms to see if you can spot frequent occurrence of keywords used, create a new field in your table for context, and make global updates to your database table. For instance, words such as careers, hr, human resources, employment, positions, vacancies, appointments, etc can all be grouped under employment.

Analysis of search term errors
As part of the classification exercise described in Search Context Analysis, it is also advisable to classify errors so that you can device techniques to minimise them. The nature of errors might be various but common types are hitting the search button without entering a keyword, misspellings, mistakes in using notations such as + and -. After examination of frequently occurring errors you may figure out some ways of dealing with them. For instance, modifying the search help text, associating frequently misspelled words to their content items, etc.

Analysis of search terms by user segments
Another analysis may involve examining search terms by major user segments. For instance, the searches conducted by internal users, business partners, major customers due to their familiarity with your services and website may differ significantly from that of a prospect, a job seeker or a researcher. Based on their IP addresses, well known user segments might be easily isolated to conduct this analysis. Service priorities placed on user segments may make you favour certain groups - for example, customers or business partners – and you may want to optimise your search engine performance for these segments first.

The effort pays off

As you can see, search term analysis requires a reasonable amount of effort. The first three techniques can be performed relatively quickly, but you should allow more time for the last three. In most cases, a few days would be enough for a comprehensive analysis.

The frequency with which you repeat the analysis really depends on the outcomes of your first analysis and the role your website plays in your business. If you discovered shortcomings and actioned them, you should repeat the analysis to determine how effective the changes were. Although for most businesses a bi-monthly or quarterly analysis would suffice, you should do it more frequently if you are interested in trends in search usage, if your website is an integral part of your business, or if it integrates with your core processes such as sales and fulfilment.

Considering a large proportion of online users resort to keyword search at some stage during their visits, without search keyword analysis, your website analysis would be incomplete. Findings from search keyword analysis will help you improve the online customer experience.