I've learned a long time ago in hockey about the concept of "playing the percentages". What this means is that a certain action or play will have a particular success factor over others given the situation. For instance, in a power play when a team is shorthanded one person, one of the best (high percentage) ways to get the puck into the neutral zone and out of your zone is to shoot it off the boards or the glass. The only risk here is having an unpredictable bounce off the glass and the puck stays in the zone. The alternative is to shoot it out in the center through a gap and risk the puck staying in the zone with the opponent quickly covering any gap. One last alternative is to actually carry the puck out of the zone in risk of shorthanding your defense further if the puck-carrier makes a mistake.
So how does this apply to the field of UX?
Well, if a certain play has a specific percentage of success, something that can be measured, it can also be applied to making design decisions - could it not?
The reason why I'm exploring this concept is because we're currently developing a U.I. Pattern Library. (Here's a case study.) And because we're rapidly developing these patterns, there will be some that need to be used immediately and thus cannot be tested right away. So we need to explore all options for a solution to a pattern. There are indeed some patterns that are "straightforward" but then there are those requiring a little more work. And with this work, we need to figure out the best solution. How better to do this than by design by the percentages?
The fact is, most UX designers and analysts have basic knowledge to design based on past data and experiences. (e.g. We know when to use radio buttons instead of checkboxes.) The caveat is that an untested pattern may be completely wrong, especially if after testing, we find out the user behavior and expectations are mismatched. By designing by percentages, we mitigate enough of the unpredictable and also keep cognoscente of our decisions and how they came to be. (This also means that there needs to be documentation that will log these decisions.)
So how can this be used immediately? To me, it's more of a concept I keep in the back of my mind. I think of alternatives and go through a cognitive walkthrough to anticipate what the user may interpret and thus behave when interfacing with the product. However, doing just that may not be as rigorous as some might like. To make is rigorous and more apparent, documentation can accompany each decision made along with alternatives and why they were not chosen. While it can take long to do, at least when the patterns are tested afterwards, the assumptions can be referenced and validated/invalidated.
In the end, this concept is not 100% correct though it does give you progress because you're spending time worthy on exploring the pattern probabilities. To adopt Bill Buxton's axiom that you can only design the framework and not the user behavior, design by percentages can at least get you one step closer to designing an effective framework that gets you eventually closer to influence user behavior.