As consultant, I have the privilege of learning from a lot of clients. There are different ways of writing software, and I get to see a lot of them.
One pattern that I see frequently are units of code whose size makes them a problem. Hatton’s frequently referenced study founds that the rate of defects climbs rapidly when a module is larger than 300 lines. (IEEE Software. Les Hatton. “Re-examining the Defect-Density versus Component Size Distribution”. March/April 1997.) That rate is per line of code, so a 400 line procedure has something like four times as many errors as a two hundred line procedure.
Read the rest on the Optimal BI blog.