Change only one variable at a time when testing improvements

Change only one variable at a time when testing improvements so that observed effects can be attributed to specific causes rather than confounded with simultaneous changes.