3D concrete printing is an additive manufacturing process in which elongated beads are assembled in layers to form 3D parts. In cementitious materials, water evaporation as well as the setting of the material results in a volumetric shrinkage of the printed structures. This imposed shrinkage strain, commonly refereed to as eigenstrain, is the source of residual progressive stress evolution within the printed parts. Consequently, the accumulation of stress during fabrication and drying may induce cracks, buckling, and surface defects in the final product. Evaluating and modeling the effect of shrinkage on the printed beads is therefore essential to optimize the machine path and process parameters in order to mitigate such issues and guarantee the integrity of the printed parts. In particular, this paper focuses on the development of a compensation strategy such that the final geometry correctly approximates the target geometry. The proposed approach relies on the experimental characterization of the displacement field and hence the total strain by using Digital Image Correlation performed on in-situ imaging of the process. The history of the eigenstrain strain (i.e., shrinkage) has been measured on flat rectangular thin-walled walls, and then used in a mechanical model as an imposed strain. Resulting geometrical distortions have been validated against experiments. On this basis, a simple compensation strategy is proposed consisting in correcting the initial machine path by the opposite of the computed distortions when the structure is subjected to shrinkage. Several examples on various part geometries are presented and discussed. A fast one-dimensional mechanical model named QuadWire proposed recently is being used for numerical simulations, as the final objective is to create large database to train neural network algorithms in order to apply this compensation strategy in realtime during the printing process.