The use of non-starch polysaccharide (NSP) enzymes has increased in recent years with most of the growth coming from the market segment which uses non-viscous diets. A variety of product choices are available to the feed industry. These range from mono-component, single enzymes, to cocktails of more than one activity, to so-called ‘complex’ enzymes, displaying, in addition to the main activity, several non-targeted (i.e. quality controlled or assured) activities within one product. The relative abundance of the substrate or the number of substrates presented by a given diet may not be viewed as the sole criterion for the fitness of an enzyme solution. An argument can be made that NSP enzymes should not be considered as classical digestive enzymes and any response may not be regarded simply as a function of the extent of in vitro or in vivo substrate hydrolysis. The idea of having additional non-NSP or NSP activities to ‘strengthen’ a xylanase response, as well as the notion ‘complex diet needs a complex enzyme’ appear to lack sufficient scientific backing. Measures based on alternative responses e.g. gut morphology, nutrient or energy digestibility, gut-flora and its metabolites or fermentation profiles, are useful in developing a wider understanding of the phenomenon but require careful interpretation as the stand-alone criteria of usefulness of an NSP enzyme. Performance data is always the ultimate judge of the efficacy of a feed enzyme. This review addresses the practical question of selecting an NSP enzyme. The aim is to discuss the data supporting some of the common views held in the industry today, and how these views significantly influence the process of selecting an NSP enzyme for commercial in-feed application.