This work addresses a knowledge gap in meso-scale modelling of concrete – the effect of selected computational ITZ thickness on predicted macroscopic behaviour. A novel approach for models' generation with arbitrarily ITZ thickness is developed, offering lower computational cost with accurate ITZ representation. Models are validated by comparison with published experimental results and verified by comparison with published simulation results. Systematic parametric study is performed to clarify the effect of ITZ thickness selection in combination with other meso-structure parameters on the predicted concrete behaviour. Results indicate that tensile strength is increasingly underestimated, up to 25%, as the ITZ thickness increases from its average physical value of 0.05 mm to the commonly used 1 mm. Correlations between selected ITZ thickness and aggregate shape, maximum aggregate size, aggregate volume fraction and ITZ tensile strength are also studied, providing guidance for future research with meso-scale models.