I will explain a framework to study the non-planar corrections in N=4 SYM by using the integrabillity. Specifically I will talk about the first non-planar correction to the four-point functions of BPS operators, which correspond to a four-punctured torus on the string-theory side. The method bears some resemblance to the sewing construction of the 2d CFT: One first decomposes the worldsheet into hexagonal patches and then glues them together by summing over all possible states that appear as the intermediate states. I will also comment on similarities to and differences from the standard string perturbation theory.