Ideally, products would have no learning curve: Users would walk up to them for the very first time and achieve instant mastery. In practice, however, all applications and services, no matter how simple, will display a learning curve.
Learnability and usability are not mutually exclusive. First, decide which is the most important; then attack both with vigor. Ease of learning automatically coming at the expense of ease of use is a myth.
DIY - Case Study: Ashlar-Vellum Graphite
Rent yourself a copy of the high-end CAD/CAM program from Ashlar-Vellum called Graphite. Notice how it took you less than 10 minutes to learn how to do productive work. Then try doing the same thing with the leading CAD/CAM brand. Notice how after six weeks you’re still just staring at the screen, wondering what to do. (You can skip the other leading brand, but do rent and learn from Graphite. It will change the way you design.)
How do you decide whether learnability or usability is most important? The first thing you must do is identify frequency of use: Are you working on a product or service that will be used only once or infrequently, or is it one that will be used habitually? If it’s single-use, the answer is clear: Learnability. If someone will use this every day, eight hours a day for the rest of his or her life, the answer is equally clear: Usability.
Next, who is the buyer? If the person who will use it habitually will also make the buying decision, a product’s reputation for learnability may be a key factor in making the sale. That’s why you want to identify the most important of the two, then attack both.
Much usability testing involves running a series of tests at regular intervals with your spending only 20 minutes to an hour with each subject you recruit. You end up knowing everything about the initial learning curve and nothing about either the long-term curve or the end-state level of productivity.
If you are working on an application that will be used habitually, go about it entirely differently: Work with HR to hire temporary workers. Then, have them spend a week or two coming up to speed on the interface, monitoring them with time tests to see what the overall learning curve and the eventual efficiency of your interface actually prove out to be.