Gary Angel's post on Evaluating Internal Search Performance By Topic deals with an issue I am currently working on and that I heard a lot about, at Search Engine Strategies a couple of weeks ago, here in NYC.

One of my clients has a search datamart that we use to pull no results and click through rate by brand - except, there's no categorization built into this data-mart - so you can't get a readout of search terms by brand; instead, we use wildcards for the brand name and do the same for no results - but use different filters.

The problems I've had with the approach we're using for this client are:

1.  We're not using all the right terms as wildcards - we're getting a lot of junk and missing a lot of good stuff - but we still do get some information the way we're doing it.
2. Reports take a long time to run and often crash - they need to be staggered and could be automated but often need to be watched so they don't clog up the rest of the reports running.
3. No Results are one thing - but acceptable Click Through Rate for internal search results is a user defined thing - there's no real standard an alot of it may be due to the the way the search engine ranks results and takes snippet of text to display on the search result page.  As a result - I'm not sure the reports are really as helpful as some people think they are.

Gary Angel has an interesting idea that's a little different and may work better (if it can be implemented).

• Segment Definition (missing from our client's data) - categorizing the search terms the visitor types in to the search box on your site.
• Categorize the Search Results being clicked on - an OnClick handler to each Search Result. This handler should make a dynamic call to your measurement solution (or http server if you’re log-based) – and register the Search Category Event. One benefit to adding OnClick handlers to Search is that you can also record the position of the clicked Search item.

If we had both segment definition and search results categorization - and this was set up to be brand focused, the reports I would be generating could be both much easier to generate and much more useful - in my opinion.

Another idea I wanted to bring up from Search Engine Strategies - datamining your search logs and categorizing the data based on the overall intent searchers have when they come to your site is worthwhile information.  Once you have that information - you can look at your search landing pages to see if they match the intent of the visitors to your site - and if not, change the pages.

I'm glad I read Gary's post tonight as it gave me some ideas of what to do next with internal search projects.