Universal swill-army-knife for a numerical or categorical data analysis and rapid prototyping. Compared to domain-specific languages (eg. R, Matlab, SQL) - it adds the simplicity, modularity and universality of the backend python. Powerful features as a OOP friendly data objects, smooth I/O features incl. sql, csv, xlsx..., easy slicing dicing filtering, built-in or custom functions, computation power of scipy/numpy, instant visualization with matplotlib..., or an easy integration into any python deployment framework (e.g. django) makes the pandas simply rock. ... The only disadvantages: some builds could be unstable for a production, some computations are limited by available RAM, used OS (e.g. 32-bit systems) or 1 threaded computation due to python GIL... Nevertheless this is highly recommended tool which can be complementary (sometimes even replace) some of its more expensive commercial BI tools.
I just hope that more of R - statistical magics will by directly integrated in the near future.