Abstract Paddy rice fields are kept inundated during most of the growing period. This requirement is challenging to achieve because of the lack of suitable technologies to detect rapidly percolation prone zones within these fields. The objective of this study was to evaluate a methodology to identify water leakage areas to support precision soil–water management at a within-field level. Therefore, a Floating Sensing System (FloSSy) was designed to record the soil apparent electrical conductivity (ECa) of a paddy field both under dry and inundated conditions using the electromagnetic induction sensor EM38. Comparison of ECa data sets obtained under inundated and dry conditions showed that the ECa measurements under inundated condition (ECa-i) were more strongly related to soil properties due to the absence of variability in soil moisture and the increased stability of the floating sensing platform. Therefore, we proceeded with the ECa-i measurements and grouped them into two classes using a fuzzy k-means classification method. These classes showed significant differences in water infiltration: lower ECa values represented a higher infiltration rate and vice versa. This effect was attributed to differences in soil texture, more specifically the sand content, and its effect on water retention. It was concluded that an ECa-i survey with FloSSy allowed the detection of soil heterogeneity linked to downward water fluxes which has a potential to support precision soil–water management in inundated fields.