The best way to improve a complex system is by trial and error. Don't just take my word for it; watch this Ted Talk presentation by Tim Harford, or this TEDxEutropolis talk by Paul Ilske. Of course, if you use trial and error then you have to accept failure. Not all of your changes will turn out to be improvements. It is equally important to learn what NOT to do as to learn what you SHOULD do. In fact, Thomas Edison once said when he was reminded how many failed experiments it took him to build a functioning light bulb: "I haven't failed, I now know over 300 ways NOT to build a light bulb".
Too many organisations have an explicit or implicit dislike for failure. In business development organisations this is even more common than in software development organisations. This fear of failure leads to stagnation and playing it safe and I think is one of the primary reasons great start-ups lose that dynamic and explosive feeling. At the beginning you have nothing to lose and therefore experimentation (and therefore failure) is more acceptable. But as the business grows so the organisation becomes more afraid of failure and more conservative. Trial and error are no longer acceptable. It is no longer enough to say "I have this great idea I'd like to try out", you now have to provide days of analysis work, risk assessments, etc. etc. and above all be absolutely sure that it will not fail!
Instead of focusing on how to prevent failure, organisations should focus on ways to evaluate the results of changes. We need to be able to assess if a change was an improvement or not. If it was an improvement we want to keep doing it, or even better build upon it. If it wasn't an improvement then we want to go back to the previous successful state. This process should be simple and there should be tools and metrics in place that allow us to make this evaluation. There needs also to be a culture that allows us to make changes and more importantly to reverse our changes in a positive way. There should be no negative stigma in discovering a path of development we shouldn't take, be it in our process or our product.
I think this is one of the drivers as to why Fed Ex days, innovation days, and other such initiatives work so well; they allow experimentation. And through experimentation we learn and improve, because we are hard-wired to learn this way. We are hard-wired to learn through trial and error how to walk and talk. When children play they are constantly experimenting. So why, when we become adults is experimentation suddenly taboo? Why must every change succeed? Food for thought I think...