In Great by Choice, the most recent book by famed author Jim Collins, he talks about 10x companies (companies that outperform the market by 10x) and how they do a better job shooting bullets, as opposed to cannonballs, for new opportunities. Think about the cost of bullets and guns vs the cost of cannonballs and cannons — bullets are significantly cheaper than cannonballs.
Most startups come up with a new product idea and fire a giant cannonball at it without knowing if they are pointing in the right direction and actually have a target. Instead, they should use bullets and iterate quickly to try and find the target. Once the target has been squarely hit by the bullet, doubling down and loading the resources to fire the cannonball at the same target becomes a much more effective strategy.
Startups should shoot bullets to find a cannonball opportunity. This is another way of saying start with a minimum viable product and iterate quickly via customer driven development — a core part of the lean startup movement. The next time a startup builds a full product and launches it with significant resources, and it doesn’t work out, think to yourself that they should have fired bullets until they hit the target, instead of lobbing a cannonball out there right off the bat.
What else? What do you think about startups shooting bullets to find a cannonball opportunity?
David Cummings on Startups 
So bullets are analogous to experimentation? Can’t agree more that in a knowledge base economy the need for more experiments and creativity. So how do you build in room for experimentation at your companies? There is always an existing customer calling for help/fixes, a development roadmap with features planned but not yet built, and new customers to be pursued.
I have heard the argument for actually overstaffing engineering talent as a valid way of solving for this problem. Obviously assumes one can afford to overstaff, and perhaps overstaff is even the wrong word, but more or less the concept is to staff more talent than current business forecasts project. The bench of staff that are “slow” should be incentivized to pursue business pertinent features, enhancements, fixes to existing products and through a natural selection process harvest the great ideas/fixes/features that survive experimentation.
Obviously an incomplete thought, not fully realized by this comment, but interested in the subject of your post, per the usual.