The best thing to do is to define tradeoff ranges and set the deadline, so that developers could try an optimistic path, learn what’s wrong with it halfway, and

The best thing to do is to define tradeoff ranges and set the deadline, so that developers could try an optimistic path, learn what’s wrong with it halfway, and then choose a tradeoff that both seems not that bad enough and, when it fails, leaves time to rethink and use another tradeoff which is even good now, in a sense. Because that’s what we do either way. Round numbers from ass make no science to plan and prioritize upon.