My point was that sentence guidelines should generally be data-driven, the dependent variable being the crime rate for a specific offense. I did not mention edge cases, judicial discretion, capital offenses, recidivism, rehabilitation, and other gnarly stuff.
Does in your example crimes with lower rate of catching criminals will end with a very high sentences and few caught will get an unreasonable jail times? "You have pirated a song, an infinity in jail is Your sentence"
"Data-driven" is a reasonable approach, but what data should be allowed to drive? Choosing what kind of data to use and making relevant conclusions from it is the hard part.
Anyways, the data is in. Putting more people in prison for longer doesn't work. The problem though is that in the short term reversing this policy without any other policy changes may not work either. People aren't as simple as that.