Measuring the effectiveness of agri-environment schemes depends on scheme type, taxon and landscape. Here we show how spatial scale, i.e. studied transect, field or farm level, and controlling for yield loss, can drastically change the evaluation of biodiversity benefits of on-field (organic farming) vs. off-field (flower strips) schemes. Transects may lead to misleading evaluations, because flower strips, covering only 5% of conventional fields, support less bees than large organic fields; but if their 20% yield loss is considered to compare identical yield levels, 80 ha conventional plus 20 ha flower strip farming promotes more bees than 100 ha organic farming.