Prototyping That’s Less Prone to Failure
16 December, 2015 / Articles
When I was a young industrial designer at the erstwhile Sun Microsystems (now owned by Oracle), I was working on the design of a large server. The product had modules that could be removed and replaced quickly by system administrators if something went wrong. We wanted to find the most intuitive design for the rotating latches on either side of each module. Since engineering was on a tight schedule and physical prototypes were months away from being ready, we elected to do “paper prototyping,” where we showed drawings to users of the modules with different latch designs.
The results came back moderately in favor of the design that I preferred, which involved the latches turning inward like sink faucet handles (appealing because the module was visually symmetrical). But when the physical prototypes came in months later and we put them in the hands of real users, my mistake became apparent.
Our paper test had been missing a critical element: the latches were rotated using a screwdriver. When you put a screwdriver in someone’s hand, their brain flips to the familiar “lefty-loosey/righty-tighty” mental model. But the “faucet handle” approach meant that one latch was turned counterclockwise to release it, and the other went clockwise — and everyone tried turning that one the wrong way.
The error was obvious in hindsight, but unfortunately it was too late to change the design, as this little detail had a big cascade effect on how the rest of the hardware was designed. While not a failure of catastrophic proportions, we were stuck with it, as were the poor system administrators who’d have to live with this inconvenience for years to come.
I was reminded of this painful memory recently, while reading Jeffrey Rothfeder’s Driving Honda, which examines the company’s innovation culture. Honda’s emphasis on realism could have prevented my prototyping error — and it was so simple. Honda’s principles are known as “the three realities”, or sangen shugi:
Gen-ba: Physically go to the real location where the activity or problem is occurring. This could be a server room, a factory floor, a living room, inside a car on a summer road-trip – wherever you must go to gain firsthand knowledge.
Gen-butsu: Once at the location, carefully observe the conditions and begin to formulate a decision or recommendation.
Gen-jitsu: Support your decisions with data and information collected at the real location.
These “three realities” are not unique to Honda. But according to Rothfeder, many companies, such as Toyota and other Six Sigma practitioners, place undue weight on the last reality, gen-jitsu, because their focus is on data-driven process efficiency. But he observes that Honda sees gen-ba, or going to the actual location where a problem is occurring, as the most important because it ensures insights are grounded in reality, it encourages empathy, and it avoids the detachment that can come from a reliance on data.
It’s the old garbage-in/garbage-out story: if your experiment isn’t grounded in enough reality, you’ll be deluding yourself with data that has the air of truthiness but is in fact wrong. In thinking about my own latch design mistake, the failure lay in the paper prototype not giving a sufficiently realistic simulation of the actual situation — we just asked people to imagine how they’d use the latch. If we’d handed them a screwdriver and had them pretend to operate the latch, the results may well have been different.
It’s always good practice to test early and often with rough prototypes, whether they are paper-based, 3D-printed, or simple wireframes of a user interface. These methods let you answer the fundamental questions early and cheaply — before you’ve locked yourself into a design direction. But innovators should keep some rules of thumb in mind:
- All prototypes have limitations and ways in which they are going to bias users and the results. Be aware of what those are, and use prototyping methods that are geared to what you need to learn while minimizing distracting inaccuracies that could throw off results.
- As soon as you have more realistic prototypes, double-check the findings from the previous tests and see if they still hold up. For example, re-check usability of navigation that worked well with wireframes once you have polished visual design of an interface. Seemingly small visual changes can break what was working before.
- Usage context matters. It can be tempting to do quick-and-dirty testing in cafes or other locations where it’s easy to grab a sample of people. But unless your product really will be used in those locations, you should do another test in the place where your product will likely be used.
In short, innovators need to move quickly, iterate, and create rough and ready prototypes to test their ideas. And this doesn’t mean losing touch with reality. Shortcutting on realism will shortchange your work in the long run.
