Timber extraction threatens a vast area of tropical ecosystems, making it vital to design productive harvesting operations that limit biodiversity declines. Contrasting management options span a continuum from less-intensive, land-sharing logging applied over a larger area to land-sparing operations that combine intensive harvesting with the preservation of old-growth forest. Combining company-reported extraction rates with dung beetle surveys along an Amazonian logging gradient, we explore how individual species' abundances, geometric mean population sizes, functional diversity, and trait characteristics vary across simulated logging concessions and production targets. We substantially extend previous studies by evaluating 8000 mixed-harvesting scenarios and by assessing the profitability of contrasting practices. Simply maximising old-growth protection delivers the highest species' abundances and population sizes for species negatively affected by logging. Maximising old-growth also supports communities with a functional trait dissimilarity (FDis, RaoQ) and functional structure of nesting guilds, biomass, pronotum volume, front leg area, and front:back leg ratio traits that closely resembles old-growth forest. Functional evenness (FEve), richness (FRic), and divergence (FDiv) did not vary across logging strategies. Some 3 % of mixed approaches outperform extreme sparing (which maximises old-growth retention through intensive logging) but still involve substantial sparing, enabled by intensified logging elsewhere. However more-extensive business-as-usual harvesting is up to 90 % more profitable than extreme sparing, suggesting active policy mechanisms, standards, or regulations would be needed to make spatially-concentrated logging operations (which benefit biodiversity) more commercially attractive. Old-growth sparing appears key to limiting biodiversity declines within tropical timber concessions, but would require payments to compensate for reduced profits.