An analysis of failures to produce in several fields showed this fault:

Example: A cook can cook 30 dishes of various types successfully. Instead of retaining these as they are and seeking on the side to create or find new dishes, the cook experiments with and changes her 30 standard dishes. The result is failed production.

Example: A musical group has 15 finished pieces of repertoire. Instead of developing totally new pieces, they rewrite their existing repertoire. The result is a failure to do good shows.

Example: An org is doing well with a standard CF letter writing campaign. This personnel is pulled off onto phones only as an experiment. The org stats crash. The correct action would be a pilot phone program using new personnel and leaving the standard actions in.

In all cases the right thing to do is maintain without variation the standard production line and if experimenting or change is to be done: pilot it on the side with people or actions that do NOT impede standard production.

There is always a better model in the research lab than there is on the production line. The only bug occurs when the incomplete and unknown model is shoved over as the standard production.

If on test and experience a new action, properly piloted, is better, then and only then is it added to the standard line.