A question was posed by a good friend and colleague a while back about testing for learnability. The discussion was quite good in terms of defining possible qualitative data to compare. However, what I'd like to know, is there a simple, quantitative way of testing for learnability?
Here's my take, and unfortunately, it's only a theory for now because I haven't practiced it:
There's one metric that hasn't been mentioned and that is time-to-task. Now, what if time-to-task can be measured several times on the same task as opposed to just once? If the user is suppose to learn about how to use an application, shouldn't the time-to-task be reduced on subsequent tries? Isn't that an indication that the user have learned how to use it? When we play video games, the second or third time around seems easier to get through a level, doesn't it? So in that sense, how do we test for it?
After determining what kind of learnability you are testing for, a possible usability session may consist of the following guidelines:
1. Have the participant perform a repeated task.
2. In the series of tasks, mix in another task that could be something completely different (or something related). This disjointedness could test the user on how well they remember to perform the previous task again.
3. Increase the time between the first time and the subsequent times of performing the repeated task.
This method can be a sure way of quantifying the learnability of a particular function or process because what will happen with a successful design is that the time-to-task should be decreased with each repeated tries. Something that is very learnable may takes significantly less time than the initial stab. If there is no difference on subsequent tries, well, you know it isn't learnable.
The idea is to reduce the learning curve (dictated by time) and with this method, the time-to-task metric can be used to its full potential.