Much effort has been spent on promoting conservation agriculture (CA) in Northern Benin to sustain the transition of cotton (Gossypium hirsutum L.) cropping systems toward agroecology. However, its limited adoption by farmers is often ascribed to initial yield penalties during the transition to CA and to trade-offs around crop biomass use. Here, we assess the effect of different CA-based cropping systems promoted in the region on water productivity and cotton yield in a three-year cotton/maize (Zea mays L.) crop rotation during the initial transition phase to CA. Three CA options were assessed combining different levels of soil disturbance and cover, and introducing cover crops to alleviate the biomass trade-offs. Direct seeding (DS), strip tillage (ST), and direct seeding mulched-based cropping systems (DMC) were compared with conventional tillage (CT) from 2017 to 2019 under a dominant soil type in the region, Haplic Lixisols. Two legume species, Stylosanthes guianensis (Aubl.) Sw. and Crotalaria retusa L. were grown as cover crops with maize under ST and DMC. The experiment followed a randomized block design comprising six replicates. After 2–3 years of DMC, the cotton yield advantage with respect to CT increased from 5¿% to 7¿%. Cotton yield penalties of respectively 11¿% in 2018 and 26¿% in 2019 were found for DS. ST treatment went from a yield advantage of 8¿% in 2017 to a yield penalty of 20¿% in 2019. The DMC and CT treatments gave similar and highest boll weights compared to the ST and DS treatments. The treatments had no significant difference regarding the number of bolls per plant. Soil water storage in the upper 30¿cm depth and water use efficiency (WUE) were the highest in the plots with the DMC treatment compared to CT, ST, and DS. At 28 days of active vegetative stage (between 34 and 62 days after sowing), the WUE of seed cotton was 0.11¿kg¿ha-1 mm-1 under DMC, while it was 0.08, 0.07, and 0.04¿kg¿ha-1 mm-1 under DS, CT, and ST, respectively. The per