The Gross-Pitaevskii formulation is essentially a classical treatment, very useful, but missing a great many of the properties that proper quantum systems possess. When you try and calculate the dynamics of interacting systems, this can become a problem.
The Truncated Wigner approximation addresses several of these issues. First, the indeterminate nature of states. Because you can't say for certain that there are exactly X particles in a system at any given time, as the GP formulation does, the TWA samples states immediately around the average measured value and evolves each of these, taking a combination of the resultant trajectory to calculate the expectation value.
This is useful in situations such as those observed when the particles in a BEC will decohere over time. TWA demonstrates this very nicely (though still fails to predict a number of phenomena, as always with approximations)
The TWA can be extended to include spontaneous jumps in the state and other purely quantum phenomena for which the GPA has nothing to say. As usual, this accuracy comes at the expense of ease of calculation. The corrections particularly scale at horrifying rates for anyone attempting to model them numerically, as I am.
I hope this is a help to anyone trying to understand this theory from scratch - no one seems to have explained this simply that I could find.