Sequential importance sampling (SIS) is a procedure which can be used to sample contingency tables with constraints. It proceeds by simply sampling cell entries of the contingency table sequentially and terminate at the last cell such that the final distribution is approximately uniform. I will first introduce this procedure in both statistical and polyhedral geometry view, and explain its advantages and problems. Then I will introduce the SIS procedure via conditional Poisson (CP) distribution which is used to sample zero-one contingency tables with fixed marginal sums. In this case, the procedure proceeds by sampling one column after another sequentially and terminate at the last column. I will explain both two-way and multi-way cases, and also why it performs better than the general SIS procedure when we have zero-one constraints.