I would use the river analogy but slightly different than Kostas (he sounds like he does a lot of size exclusion chromatography).
Nobody wants to be washed down the river so when they are dropped in the first thing they do is latch on to the bank and hold on.  All your various compounds, like people, have different degrees of strength so the weak ones (polar compounds) can’t hold on very long and are washed downstream quickly.  Your strong individuals (non-polar) can sometimes hang on forever.  This is where solvent strength comes into play.  Seeing as the rivers flow is constant, an increase in solvent is not an increase in flow but an additive to the river that makes everything slippery and more difficult to hang on to.  Eventually, with enough solvent introduced into the river nobody can hold on and they are all washed away.  
This also explains why if you are only dropping one individual in the river you don’t need to run a gradient.  You get to know this individual’s strength and can introduce just the right amount of your slippery additive such that he holds on as long as you want him to. When he exits he is easy to detect because he is all alone. 
However, when you drop a whole crowd in the river at the same time (a dirty sample) you need to run a gradient.  You add very little solvent such that everybody gets a good grip immediately.  Then as you start ramping your slippery additive the individuals start loosing their grip and washing away one at a time such that when the guy you are looking for lets go he doesn’t do it at the same time as anyone else and also exits alone and easy to detect.
This also explains the use of different rivers (columns) when analyzing various people.  Your C18 has a whole bunch of handholds for people to grab on to so if you are analyzing really strong people you might want to use one that doesn’t have so many.
