On the analysis of big terrible data