It is usually not practical for invertebrates to be comprehensively included in biodiversity surveys that underpin conservation planning, and so a representative subset of taxa needs to be selected. One approach to representativeness is to select taxa whose patterns of richness and composition are most strongly correlated with those of total invertebrates (i.e. all taxa combined). However, if different groups show very different distribution patterns then ‘total diversity’ cannot be considered as representative of the diversity of invertebrate taxa, and so an alternative approach to achieving representativeness is to base selection on complementarity (i.e. representing the full range of distribution patterns shown by different taxa). We use data on 12 invertebrate families (comprising ants, beetles, flies and spiders) sampled using pitfall traps across 78 sites in a tropical savanna landscape of northern Australia to identify a subset of target taxa (families) to represent their diversity patterns. We use a simple scoring system that incorporates both survey practicality and biological representativeness to compare selection of taxa based on (1) representing ‘total diversity’ and (2) representativeness through complementarity (‘complementary diversity’). Congruence among taxa in terms of both species richness and composition was generally low (ρ < 0.5), suggesting that taxa are poorly representative of each other and thus a complementary approach is required for target taxa selection. The taxa that scored highest in representing ‘complementary diversity’ were very different to those representing ‘total diversity’. To our knowledge, this is the first time that invertebrate representativeness based on ‘total diversity’ and ‘complementary diversity’ have been directly compared. The selected target taxa are specific to our study system, but our simple method for selecting representative invertebrate taxa for conservation planning is widely applicable, including for biodiversity monitoring and environmental impact assessment.