A colleague was working with a company where the "order configuration" function was the key bottleneck in their business. The configuration function (~20 people) consistently delivered their projects late and/or under-spec, delaying the remaining work. In working with this client, they cut back the amount of work released into the system, and their performance went up: they moved above 75% on time delivery.
But when they hit these wonderful milestones, they thought their problems has been solved and started releasing more work into the system. The system performance immediately dropped down into the 25% range. Happily, they've seen the mistake and have moved back in the right directly.
Nimmy gives us a nice quote that fits this situation wonderfully, and he says it's a German proverb, Time for a paradox!
German Proverb - Who begins too much accomplishes little.
But what is too much? Queueing theory suggests that any resource loaded above 90% is going to create a vicious cycle of poor performance due to inherent variation in most systems.
The Theory of Constraints way of thinking requires that there be mechanisms to clearly define when work should be released into a system. Most organizations start work as soon as a job has been accepted. While this might sound like a good idea, it invariably creates the situation where there is far too much to do and orders end up late. This applies to both manufacturing and project management, and is counter-intuitive for people until they see it in action.
Rather than focusing on what you start, focus on completing what you've started.