Conservation agriculture is widely promoted to reduce soil degradation, restore and maintain soil health, enhance crop productivity, and mitigate greenhouse gas emissions. This generally entails three core practices: reduced soil disturbance, permanent organic soil cover, and crop rotation. However, the universal applicability of these practices across diverse biophysical and socioeconomic contexts is under debate due to inconsistent agronomic performance and practical challenges associated with implementing all three practices simultaneously. To better understand the associated biophysical dynamic, we evaluated changes in soil-plant-water relationships and crop yields under five practices: no-till without residue (NT), reduced tillage without residue (RT), no-till with residue (NT+RR), reduced tillage with residue (RT+RR) and conventional tillage with residue retention (CT+RR), each compared with conventional tillage without residue retention (CT), using a global meta-analysis of observations from 338 studies across 361 experimental sites. Overall, yield declined by 6.1% (p = 0.024) under NT. However, yield reductions diminished with increasing tillage intensity (from NT to CT) and residue retention (p = 0.041). Sensitivity analysis revealed that yield reductions under NT are likely driven by compaction-driven adverse changes in soil hydraulic and mechanical properties limiting water movement (infiltration, redistribution and drainage), retention, and availability to crops, and root growth. Specifically, NT, irrespective of residue management, reduced soil near-saturated hydraulic conductivity – a property governing soil water replenishment, redistribution and drainage – by 26.9% (p = 0.008), increased soil penetration resistance by 29.4% (p < 0.001), and changed the pore size distribution, resulting in smaller air-capacities and larger wilting points. When reduced tillage and residue retention treatments were combined (RT+RR), yield variability was more strongly associated with changes in soil organic carbon and saturated hydraulic conductivity than water retention, and penetration resistance. Yield responses varied with local climate and soil type. NT increased yields in semi-dry climates (aridity index: 0.3-0.65) by 16.3% (p = 0.004). In contrast, NT reduced yield in humid regions (-7.2%, p < 0.001) as well as in dry regions (-8.3%, p = 0.038) where irrigated agriculture is likely to dominate. These yield responses by climate context closely mirrored the observed differences in saturated hydraulic conductivity. Yield penalties were generally greatest in clayey soils (e.g., under RT -19.3%, p = 0.047) and consistently diminished toward sandy soils, both under NT and RT. These findings highlight the need for context-specific implementation of conservation agriculture to achieve balanced agronomic and environmental benefits.