A few teams in a 500 person development group at IDX Systems (now GE Healthcare) generated production code at five times the industry average, while most of the teams who executed SCRUM, only doubled productivity over industry average. One of the problems in the large organization is that it was culturally prone to a team size of about 15 people and there was a lot of internal resistance to reducing team size. I think this is the primary reason only a few teams moved into high performance mode. The hyperproductive teams would always split into subgroups of 7 or less, while the poorer performing teams insisted on working as a group of 15.
Quantitative Software Management selected data from 491 projects of roughly equal size. Six person teams averaged 11 months to get the project done. 10 person teams delay the project by 6 months. Any team over 7 in size should be split up into multiple SCRUM teams.
Average cost per function point across over 1000 projects in Rubin's Worldwide Benchmark database is $2970. For teams of size 7, the average cost was $566 per function point. Most companies productivity is (by definition) the industry average in function point analyses done by Software Productivity Research. We may spend about 6 times the necessary cost for each project we do on the average.
Rubin, Howard (Ed.) A Metrics View of Software Engineering Performance Across Industries. IT Metrics Strategies V:9:3, September 1999.