We have reached some conclusions after reading and organizing more than a hundred papers on scaling up inductive algorithms. Much work has been done, and the establishment of KDD as an coherent eld of study seemingly has accelerated the production of relevant results.