Datasets with a large amount of noisy data are quite com-mon in real-world classification problems. Robustness is an important characteristic of state-of-the-art classifiers that use error minimization techniques, thus requiring a long time to converge. This paper presents ClusWiSARD, a clustering customization of the WiSARD weightless neu-ral network model, applied to credit analysis, a non-trivial real-world prob-lem. Experimental evidence show that ClusWiSARD is very competitive with Support Vector Machine (SVM) w.r.t. accuracy, with the difference of being capable of online learning. Nonetheless, it outperforms SVM in both training time, being two orders of magnitude faster, and test time, being slightly faster.