Finding a balance between the agroecological benefits of in-field crop diversification and the associated management demands, while maintaining expected production levels, is essential for making crop diversity work for farmers. The aim of this study was to find a workable resolution of diversity within the context of an on-station organic arable cropping systems experiment in the Netherlands. The experiment tested a gradient of crop diversity treatments from sole-crop references to strip cropping (3 m x 54 m strips sown in adjacent crop pairs of varying complexity) and pixel cropping (0.25 m2 plots each sown with one out of six total crops and arranged in 7.5 m x 12 m grids). In these treatments we assessed the performance of multiple agroecosystem service (AES) indicators (soil fertility, crop yield and quality, weed cover and diversity, and natural enemy activity density) for three focal crops (cabbage, wheat, and potato) using three years of field data and a three-part analysis. First, we used linear mixed models to assess the effects of each diversification treatment on the AES indicators. We found no clear indication that one treatment performed better than the rest across AES indicators. Second, we developed a novel method for quantifying the temporal, spatial, and genetic structural diversity of the tested treatments into compound diversity scores, and used these scores to analyze response relationships between increasing in-field diversity and AES delivery. Here we found that increasing compound diversity had a positive effect on the indicators weed species diversity and natural enemy activity density. For production indicators, we observed an inflection point between the most diverse strip cropping treatment and the pixel cropping treatment, with pixel cropping performing notably poorly. Third, we used a multivariate analysis approach to assess the contribution of temporal, spatial, and genetic diversity to AES delivery, but found no clear effects of individual diversity dimensions on AES delivery. ...