A sediment fingerprinting approach was applied to identify dominant sediment sources in an area where soil conservation measures (i.e. terracing) had been carried out on steep, intensively cultivated lands but the outcome was unknown. The wider purpose was to provide scientific evidence to inform decisions on where erosion control and sediment mitigation strategies could be further targeted. Geochemical fingerprints were used to quantify sediment contributions from three potential sources, i.e. surface soil under cropland and woodland land use, and channel banks, in a managed small catchment in the Upper Yangtze River basin in southwestern China. In parallel, artificial mixtures with known source proportions were evaluated to examine the effects of grain size selection (<125 µm and < 63 µm) on the accuracy of modeled source contributions. Source apportionment results suggest that materials originating from incised and actively eroding channel banks were the most important source of sediment, which contribute over 80% of sediment to the catchment outlet. Sediment inputs from cropland (10–20%) and woodland (<10%) areas as a result of surface erosion were less important, since effective soil conservation measures have been implemented in this catchment. Although apportionment of sampled sediment provided comparable results for both coarser (<125 µm) and fine (<63 µm) size fractions, the artificial mixture results indicated that unmixing the coarse fraction alone could yield poor agreement between modeled source contributions and actual source proportions. The mean absolute error (MAE) for the coarse fraction mixtures ranged between 8.8% and 19.6%, with a mean of 13.6%, compared to the values of 4.0–7.4%, with a mean of 5.2% for the fine fractions. The results of this study highlight that channel bank materials constitute a significant fraction of suspended sediment exports in a heavily managed agricultural catchment, suggesting that future conservation works should be focused on drivers of erosion from this particular source type. Herein, it is surmised that reworking of legacy valley fill deposits is tempering the downstream benefits (e.g. reduced siltation) of recent upslope soil conservation, an important message for policy makers. The findings of this work also emphasize the methodological need to take account of potential uncertainties associate with source apportionments when using specific particle size fractions in fingerprinting studies.