So, imagine you have a ship bearing down on you and you have a certain amount of gunpowder. That ship's coming and you turn and you take all your gun powder and you fire it at that ship, and that cannonball sails out there and it misses. You turn, you're out of gunpowder, and here comes the ship. You're in trouble. But suppose instead you took a little bit of gunpowder and you put it in a bullet and you fired it and it missed, but it's kind of directionally right. So, you take another bullet and you fire, now you're 20 degrees off. You fire it again and then—ping—you hear the side of the ship coming and you know you have a calibrated line of sight. Once you know you have a calibrated line of sight, you take your gunpowder, you put it in a cannonball, and you fire it on that calibrated line of sight, and you hit the ship.
We found out the reason that the comparison companies did not do well compared to their paired match was that they either (a) didn't fire enough bullets as a hedge against uncertainty and defined new things that would work; or (b) they would fire big, uncalibrated cannonballs in an attempt to look bold and visionary and then miss; or (c) once they had calibration, they lacked the guts to go big. See, because the only way to get big results is to do big things at some point. The question is, what are the right big things? The right big things are the things you've empirically validated. So, you fire bullets, you validate, then you go big—bullets, then cannonballs—it's both.