Despite a wealth of case-specific insights from agricultural adoption studies, we lack systematic evidence on which technology characteristics matter for adoption across different innovation contexts. We synthesise the results of 304 quantitative farm-level adoption studies for a wide range of agricultural innovations across more than 60 countries using multi-level meta regression. Our results show that land, capital and knowhow are generally more important when an innovation uses the respective factor intensively, but this effect is reduced in contexts where the factor is abundant. Our findings have implications for the design of rural development and agricultural extension programmes. Both should consider the interplay of geographic context and innovation characteristics to develop more effective sustainable intensification strategies.