We present the optimal reconstruction of the thermal Sunyaev-Zeldovich (tSZ) effect signal based on the combination of a heterogeneous dataset consisting of ACT and Planck data, with different numbers of channels, angular resolutions and noise levels. We combine both datasets using two different approaches, a Matched Multi-Filter (MMF) technique and an optimised Internal Linear Combination (ILC). We show that when applying the MMF to the combination of ACT and Planck data, the size-flux degeneracy is reduced and the signal-to-noise of clusters detected with their SZ signal improves by up to a factor of three. In the case of the optimised ILC method, we show that the tSZ map is reconstructed with a resolution of $sim 1.5$ arcmin. This is more than a factor two improvement compared with the Planck resolution, and with a very good control of noise, i.e. limited only by the intrinsic noise of the individual experiments. The combination of ACT and Planck data offers a unique opportunity to improve on the study of the pressure profiles and to study substructure in clusters through their tSZ.