Within variation, three major iron smelting furnaces were used in the Iron Age of sub‐Saharan Africa, ranging from the natural draught driven tall shaft to the forced draught powered low shaft and bowl furnaces. These furnace types are, however, mostly known from the ethnographic context. Often, archaeologists are confronted with remnants from the smelting process, forcing them to speculate on the anatomy of the extant furnaces. The presence of multiple fused tuyeres has been used to identify natural draught furnaces in the archaeological record. However, working back from smelting remains such as slag to the furnace type using physical and chemical evidence has generally proved to be ‘undoable’. Thus, when randomly selected, the chances are high that one cannot separate bowl furnace slags from those that formed in their tall or low shaft counterparts. This observation is hardly unexpected; analogous thermodynamics and thermo‐chemical reactions governed bloomery smelting irrespective of furnace type. Rehren et al. (2007) have labelled this phenomenon the ‘tyranny of system driven constraints’. In this study, we argue that the hierarchical use of statistical methods may add another layer of evidence which, when coupled to archaeological indicators, may be useful in correlating slag chemistry to furnace types used in antiquity.