Draining transitional reservoirs represents a challenge due to the spatial heterogeneity of sandstone in these units. The complex architecture of the Paleocene Barco Formation in the Catatumbo basin of Colombia is known to be derived from a collage of sub‐tidal, intertidal, and supra‐tidal plain and channel deposits. Barco lithology is therefore diverse, grading from mudstone to medium‐grained sandstone, and sporadic coal. To better understand the spatial distribution of Barco sandstones at the Sardinata field, we implemented a 3D seismic multiattribute methodology, which rendered reliable net sand thickness estimates of this reservoir. Our methodology included six steps: (1) seismic interpretation of tops and bases of three major stratigraphic cycles using hand‐tracking, since auto‐tracking functions tend to perform poorly in spatially discontinuous horizons; (2) log‐calibrated band limited seismic inversion, which rendered an acoustic impedance subvolume; (3) gamma ray and spontaneous potential log‐based net sand estimation, and further manual contouring; (4) hypothesis formulation through visual exploration of attributes (5) calculation and crossplotting of over 30 3D attributes extracted from amplitude and acoustic impedance subvolumes; and (6) multivariate linear regression to derive final attribute‐to‐sand thickness transforms for each cycle. Multivariate regression of attributes was preferred over non‐linear methods such as neural networks, for simplicity and speed of implementation. Results of this methodology included predictive net sand thickness maps for each stratigraphic cycle analyzed. The accuracy of our maps was confronted against real estimates at well locations, rendering values within few feet of each other. Attribute‐based maps provided added confidence in proposed locations of upcoming production wells.