Abstract The aim of this study was to identify a practical and defensible set of ecosystem attributes to form the basis of natural variability benchmarks for natural resource managers needing to determine the status of patch-scale species-level biodiversity within woodland and forest ecosystems. We used a form of multicriteria analysis (the analytic hierarchy process, AHP) to record and analyse the knowledge and opinions of 31 Australian ecologists on those ecosystem attributes considered most important as biodiversity surrogates, and those that were considered most feasible to assess. From a pool of 13 landscape context attributes and 62 vegetation condition attributes, practical and defensible attribute sets were identified based on AHP importance and feasibility weights and associated statistical analyses. Experts considered that, on average, landscape context attributes should contribute approximately one-third (0.36) to an assessment of within-vegetation-type species-level biodiversity status, and vegetation condition attributes should contribute the remaining two-thirds (0.64). Our analyses did, however, find a correlation between these importance weights and the spatial scales at which experts worked. The landscape context attributes: patch size, distance to nearest large patch, and connectivity were considered significantly more important than other attributes; however, connectivity received a significantly lower feasibility weight. A minimum set of 11 compositional, structural and functional vegetation condition attributes were identified: richness of native trees; cover of native trees, shrubs and perennial grasses; cover of exotic shrubs, perennial grasses, legumes and forbs; cover of organic litter; recruitment of native tree/shrub saplings; native tree health; and evidence of grazing. We compare our minimum sets of attributes with other published sets, and briefly discuss the issues surrounding the incorporation of attributes into natural variability benchmarks from which indices of terrestrial species-level biodiversity status of woodland and forest patches can be determined.