There are several different ways to test your pathfinding, but the most accurate way is to use a first click test. Use a screenshot of your homepage or a common landing page, and give your test participants a task like ‘Where would you click to locate information about x on this website?’
The most important thing to remember when writing a question for a pathfinding test is to use different language to ask the question, compared to the labels in your navigation elements. You can see an example of this in the sample test. The question asks users to locate ‘an electronic library of resource material’, rather than ‘an electronic library of publications’. This avoids false positive results, which can occur when you use language in your question that a test participant can just scan for, before actually reading or comprehending the page.
To classify a pathfinding test as a success, at least 70-80% of the clicks should be on the correct target. The sample test shows a scenario where there is definitely an issue with the label in the NYU site navigation. Only 56% of users found the correct navigation item.
In this situation, one recommendation would be to try to break out some of the navigation options to reduce ambiguity in the navigation elements - e.g. ‘News’ and ‘Publications and Resources’ or similar. When testing new changes to navigation and labels, make sure you use a variation set to get a new batch of participants who haven’t seen any of your previous tests. This keeps your data as pure and accurate as possible.