RoxAnn is a single-beam seafloor classification system using the echo-integration method to define roughness (first echo return, E1) and hardness (second, multiple, echo return, E2) parameters. The idea is that in the E1 vs. E2 diagram different seafloor types produce distinct point accumulations that can be used for classification. In natural environments, however, the data mostly display one unstructured point cloud rather than discrete point accumulations. To overcome these difficulties a new processing routine was set up based on Matlab code. This routine provides UTM coordinates, filters, necessary statistics and interpolated 2 and 3D maps of E1, E2 and the bathymetry. It performs customized color coding of the data (10,000 color bins) and interpolation after transforming E1 and E2 to one single parameter (‘E1E2’). The results include transect maps and interpolated maps colored according to E1E2. Even though the interpolated maps may be incorrect in the detail, they allow to see broad patterns that cannot easily be recognized in the transect view. In a further step, the color data form the basis for k-means fuzzy clustering. Case Study 1 is from the German Bight (SE North Sea) close to the island of Helgoland at 15 to 35 m of water depth. It is a small area (7.2 km2) characterized by bedrock and sandy areas. Ten grab samples were taken for ground truthing. The acoustic data reveal ‘hardest’ and ‘roughest’ conditions in areas characterized by stones and gravel in a fine-sandy matrix. Somewhat ‘smoother’ and ‘softer’ occurs the area characterized by bedrock and big stones. Here the grab sampler revealed only big stones, corals, rock-inhabiting plants and animals. It is either the case that in very rough areas the rocks are reflecting sound waves away from the transducer even though the reflection is strong (‘hard’) or the rock-covering biota reduce reflectivity. In the sandy area RoxAnn data clearly allow separation of unimodal fine sand (‘smooth’ and ‘soft’) and partly bimodal medium sand areas with shell detritus (‘harder’ and slightly ‘rougher’). The fuzzy cluster analysis suggests the above-mentioned 4 classes as the best solution. Case study 2 is from Potter Cove, a small fjord (7.5 km2) on King George Island (Antarctic Peninsula). All acoustic measurements and grab sampling (136 stations) were carried out from a zodiac. Potter cove is characterized by basins (>50 m water depth) with muddy sediments framed by steep slopes of bedrock with stones and gravels. Here we find the opposite of what was anticipated: The soft and smooth sediment of the basins yields harder and rougher acoustic values than the rugged rocky slopes and shallow areas. However, considering this, nonetheless a meaningful classification in 7 classes including classes indicating macroalgae was possible. It can be concluded that RoxAnn is a powerful tool for habitat mapping. However there are clearly problems when the seafloor is very rugged and/or steep. The case studies will be discussed with regard to RoxAnn performance.