So having spent some time tweaking in site search to try and get the best set of results for any given keyword. This is quite a tedious task and also much more of an art form than a science. The perception of a good search result vairies from one person to the next. So what’s next? Well ideally before even starting to optimise the in site search experience, take the time to implement search metrics. Search metrics will show the sucess or failiure of any in site search optimisation techniques and turn an art into a science. This article explores what approaches to take using Omniture sitecatalyst search metrics.
There are multiple approaches to effective in site search measurement dependant on what features are available in the Omniture account, what search technology is in use, what the purpose of the site is etc ..
If you are using a popular commerical in site search tool, you may find that integration is already available via the Genisis platform, this should take a lot of the hardwork out of your project. For products such as Endeca and Fast Enterprise search there’s some out of the box integrations.
The Genisis search platform integrations are quite specific and often product e-commerce activity focused. So even if you are using a pre integrated technolog you may find the rest of this article useful.
The two questions we want to answer, are people converting (e.g clicking on results) and are people who click on those results completing sucess events somewhere on the site? If you are e-commerce focused you may just be interested in the latter question, however this technique will help you answer both of these questions. The technique requires full access to the merchandising/conversion functionality of sitecatalyst. We will be using Products, evars and events, so get a basic understanding around these principles before reading on.
This example covers a typical publishing site integration. If you are implementing a e-commerce type solution you may want to focus your metrics around products, this is not the best approach for that model. The values are passed almost exclusively on a click event rather than on a page view. When a user types a keyword in the seach box and clicks on search or presses return the following info is passed.
product = “site search”
event 1= “insite search”
evar 1 =”keywords”
evar 2 = “filters applied”
Results page is served. Here you have the option of defining multiple sucess events based on the result which is clicked. One approach might work by assigning event 2 to result 1, event 3 to result 2 -3 and event 4 to result 4 or more. Each one of these events can then be passed on the click of the relevant result along with the product = insite and the correct evars.
Adding event 2, 3 and 4 together and dividing that by the number of searchs (event 1) in a calculated metric will give you the conversion rate for search.
The downside of this approach is depending on how granularly you want to measure conversion against placmement you could use a number of sucess events.
If you require an understanding of exactly what number result converts well. Consider using an additional evar at the time of click on result which passes the results position. Using an evar for positioning means with out using advanced sub-relations or data warehouse it’s impossible to correlate position against keyword.
What I like about the first aproach is that it limits the amount of information available with keywords and forces an approach where overall averages become the most important metric. Otherwise you may find theres just too much information to analyse.
Disagree? Or have another approach to in site search in the publishing sector? Make a comment.
Recent Comments