Virtually all tools should be analyzed to some extent. In the case of very simple tools, this may consist of a 2D strip analysis to estimate molding window and determine optimum gate sizes.

Some of the largest paybacks we have seen have been from what appears to be simple tools. Because they are so simple, they typically are nowhere approaching optimal. Cycle time reductions of 20% and greater are not unusual. Improved quality is almost always achieved as well.

A good way to evaluate the impact of using flow analysis is to analyze a number of tools (10 or more.) These should include new tools, reworks, and optimization of tools that have been in production already. 

Normally you will see some big improvements on 20-30% of the tools. About 50% will be improved less dramatically, but noticeable. The last 20-30% will not see any impact at all, In those cases however, the observation of known real world results become a valuable tool for 'calibrating' the analysis results.