Simulation of a medical-device for design-phase usability testing. (The ACT, ESC, up, and down buttons are active.)

Why it works

This HTML prototype allowed us to test terminology, information architecture, process flow, scrolling behavior, and other aspects of a medical device’s overall usability with actual patients in the target market. We overturned many myths about the usability of our existing products and terminology, and learned that several work flows were needed that didn’t previously exist. We also learned that it is possible to test both usability and the conceptual “rightness” of terminology at the same time, within some limits.

Top lessons:

• A “best practices” approach to workflow design is easily beaten by rapid prototyping and testing.
• The nuances of scrolling behavior have a huge impact on how “explorable” a list-based interface is, and therefore on adoption.
• People like some feedback, to know that the device heard them and is acting according to their wishes.
• Nobody understands “ACT” and “ESC” without some coaching, contrary to company belief. But if six of seven test subjects volunteer the same term, it is probably safe to use, contrary to company belief.
• It is important to pay attention to the “out of the box” experience, to help people get started with the device smoothly and quickly. Failure to do so endangers adoption success and customer good will, and increases support costs. Even so, new work in this area is likely to be a tough sell because “it’ll only be used once” and seems needless to the people who built the device (because they are so familiar with it).

What I’d change

The simulation could be more robust; not all of the functions available on the device are represented, most notably the ability to set and retain a value. We had to coach the test subjects to skip over these parts of the interface when they interrupted the tested tasks. And the one-page-per-screen structure of the simulation makes keeping track of the various screens and states somewhat worky; a JSON-based tree structure might make more sense in the future.

This simulation does not represent the full set of our recommendations. Unfortunately, not all that we learned made it into the final product, either.