After using Google for so long we have become quite used to an almost ‘psychic’ search. You type in something and Google will return a match, even offering suggestions if you make a spelling mistake, and you generally find what you want quite easily.

This ‘comfort zone’ we have developed can however lead to problems, as the Google search algorithm is far superior to what is used internally on many sites. When finding landing pages on client’s sites we are often forced to use search queries. Issues arise when there are a large number of products, landing pages and hence keywords. Checking all these search terms can be a time consuming endeavor. So, often it is assumed the keyword will generate a good enough search on the client’s site, landing on an appropriate enough landing page without having to check thousands of queries.

If client’s sites all employed more predictive algorithms then this methodology for generating destination URLs would be fine. Realistically many client’s site DO NOT have this feature. Take a large car parts site for example, a search for “ford hubcaps” produces a results page with 7 ‘matches’ which include 1 hub cap remover, 3 sets of hub cap screws, and 3 other items, not exactly a well focused results page! Just by adding a space between ‘hub’ and ‘cap’, the search query will return 27 results that are much more focused…

A more glaring example of this would be a search with keyword “power windows” (implying automatic, electric car windows), which returns zero results, however if this keyword is ‘reverse engineered’ and the query is instead “electric window motor” the query returns 1,422 results!

So, next time you are dealing with a client that has a large site, and you intend using search queries as your destination URLs, be sure to either check the keywords before hand or research how predictive the site search algorithm is. It may not necessarily be true that your high traffic keyword list will automatically generate good landing pages on your client’s site, if the site search is not up to scratch! In a circumstance like this your click-through-rate might be great but your conversion rate might be less than expected, due to improper ‘engineering’ of search queries related to base keywords.

What is a GOOD example of predictive search? My favourite is CDUniverse whose internal site search I think is excellent.