The minimum sand thickness for clastic reservoirs at medium burial depth between 2 500 and 3 500 meters to be predictable with current techniques is generally no more than 5 to 10 meters, while prediction of ultra-thin reservoirs with a thickness less than 5 meters remains a tough challenge. Based on the post-stack seismic data acquired and processed at different times from blocks 14 and 17 in the Andes of Ecuador, this study uses the post-stack consistent processing method driven by the structural trend surface to suppress and eliminate the interference of phase, energy, frequency and closure error on thin-bed reflection and reduce reservoir prediction uncertainty. The time-frequency attenuation, high-precision synthetic seismogram calibration method is employed to erase the accumulative time shift caused by formation absorption, accurately calibrate and analyse reflection characteristics of thin layers, and determine the minimum predominant frequency for resolving ultra-thin reservoirs. The weak reflection coefficient of thin layers is also effectively restored by using the high-resolution processing technology on post-stack broadband signals of “steady-state time-frequency-varying wavelet” without well data driving. The algorithm and workflow of facies-controlled waveform inversion are optimized based on broadband seismic waveform constraints. A series of technologies have then been developed and applied to the blocks, from which some tidal channel sand bodies of only 2 to 5 meters thick and 3 000 meters deep were successfully mapped. The drilling results of new appraisal wells and development wells verified that a prediction accuracy of at least 90 % with the methods has been reached.