After hearing a bit about this stuff from @ygolo and @Gish, I downloaded a few apps for my phone and decided to play around with this sort of mental conditioning. While progressing from n=1...2...3..., I noticed something interesting...
While I found the first couple of steps to be relatively natural, I started struggling with n=3+, and like most people sought ways to improve my performance. The first thing that came to mind was creating a mental image of the numbers in my head. That got me thinking, however, about how we humans tend to lose sight of the actual purpose of an activity by getting lost a particular detail or metric of performance. Given the purported objective of this sort of exercise (the strengthening of working memory, intellectual focus, and general intelligence), it seems counterproductive to develop those sorts of methods to enhance performance, so to speak, because the performance that is actually being enhanced is shifted from strengthening specific mental processes to achieving a higher score on a particular activity (which is meant to be a means of strengthening and evaluating said mental processes; a sort of proxy). I liken it to starting out with bicep curls, and deciding to arch your back while curling to increase the amount of weight you can lift: your biceps don't really become stronger, but you become a better curler.
In a study that ygolo linked to me, researchers observed a correlation between improvement on a particular n-back exercise and measures of general intelligence, with repeated conditioning and evaluation over time. I wonder, though, whether they would notice a statistically significant difference in that correlation between individuals depending upon their method for improvement (e.g. "I imaged a string of numbers visually" or "I developed pneumonic devices" vs. "I just kept at it and got better").