Finding patterns among tasks

As Aristotle might have said (if he had been an information architect), “one task doth not a tree make”. It’s hard to judge the effectiveness of a tree (or even part of a tree) from a single task. What we’re really looking for are patterns of behavior across several related tasks.

For example, if we have one task where participants scatter at a certain mid-level topic, the topic may need revising. However, it may also be a problem with the wording of the task. We need to be careful because it’s just a single task.

On the other hand, if we have several tasks that send participants through this part of the tree, and they keep scattering at the same topic, it’s safe to say that the topic is the problem.

In our experience, these are the most common patterns we see across tasks:

Pattern	Observation	Probable cause
Scattering under a topic	Across several tasks, participants who pass through a particular mid-level topic all pick different subtopics.	The subtopics are not clear and distinguishable.
Backtracking at a topic	Across several tasks, participants who click a particular topic (usually a mid-level topic) look at its subtopics and then back up a level.	They are misinterpreting the topic altogether. It could also mean that we are missing an important subtopic, but that’s less likely when the same thing happens across different tasks.
A topic attracts clicks when it shouldn’t	Across several tasks, participants keep choosing a topic even though it’s wrong for most of them. They may then scatter or backtrack as described above.	The topic may be an evil attractor – see Discovering evil attractors later in this chapter.
A topic doesn’t attract clicks when it should	Participants keep avoiding a certain topic, and this happens across several tasks where it’s on a correct path.	They are interpreting that topic differently than we intended. We may need to do some in-person testing to probe this further, then reword the topic according.
Giving up on a certain kind of task or in a certain section	Several similar tasks have high skip rates.	The tree is not supporting those tasks as well as it should. If we notice that people are often giving up in a particular part of our tree, that indicates that the headings in that section are not what people expect. For more, see Where they gave up earlier in this chapter.

Next: Analyzing by branches