This article concerns the realistic modelling of gas–solid reactions ubiquitous in industrial reactors. A simplistic deterministic model describing the stochastic isotropic nucleation-growth process in a powder of platelets has been strongly improved. Comparisons with a rigorous Monte-Carlo (MC) method that was used as a benchmark show that for realistic platelets, the current deterministic approach can strongly overestimate the reaction rate and underestimate the reaction duration because it considers the platelet surface to be infinite. A new deterministic model considering the finite dimensions of the platelets was developed and leads to the same results as MC simulations. Virtual experimental data were generated with this new model. Kinetic parameters were then optimised using the old and new deterministic method. The optimal curves obtained under the infinite platelet hypothesis are far from the measurements. This paper underscores that chemical kinetic parameters are only reliable if the underlying theoretical model is sound.