IMO, it means much more to everyday usage that half of blocks are much faster than 10 minutes (but the average is dragged up to 10 minutes by the long exponential tail).
To be pedantic: 63.2% of blocks are faster than 10 minutes.
Pedantry is good. I spoke imprecisely. I was thinking of the median. Statistically, exactly 50% of blocks arrive within about 6:56 m:s or less; that is >= 30.7% faster than 10 minutes, “much faster” in my book.
Although this is tangential to the context in which it was raised, I miss no opportunity to help address a common error. Most Bitcoiners get this wrong—even most regulars in Development & Technology. Most blocks take nowhere near 10 minutes to arrive; the average of the exponential distribution is almost meaningless to everyday-life usage questions of “how long will this take?”, and the exponential distribution generally defies human intuition.
People also tend not to understand that in a memoryless system, the statistically expected arrival time starts “right now”—whenever “now” is. Without checking when the last block occurred, I know that at the moment I make this post, the next block has a 25% chance of arriving within about 2:53 from now, a 50% chance of arriving within 6:56 from now, a 75% chance of arriving within 13:52 from now, and a 25% chance of arriving >= 13:52 from now. The arrival time of the last block is irrelevant.
Still, the point still stands that increasing block size does nothing to reduce block time or make bitcoin more suitable as a point of sale currency. For this, you need second layer solutions.
Increasing the block size to reduce confirmation time would be like purposely becoming obese to become a better marathon runner. Say what? It won’t work, and it’s unhealthy.