« Sizing is more of an engineering problem” said one representative of an African regulatory energy authority, evading our question related to the role of the public sector in isolated mini-grid sizing. Academia, public authorities, funding agencies, indeed tend to consider mini-grids sizing mostly through technical lenses that need to align with economic constraints. Numerous sizing software like HOMER have been fine-tuned to optimize performances, that is to decrease costs and improve technical reliability. Nonetheless, a large share of mini-grid fail after an average of six years, and sizing has been identified as a major risk. While developers benefit from elaborated sizing tools and decades of feedback, how can we understand the difficulties to size mini-grids? Based on field work in Kenya, we detail how MG developers, and especially private ones, face constraints and uncertainties regarding demand estimation, funding and regulatory aspects when sizing MG. They adapt their methods and business models to deal with those constraints and uncertainties. But as electricity demand and production capacities keep evolving, “accurate” sizing seems to be only a transitional equilibrium before or after an under or oversizing situation. While flexible sizing might be a solution, we show that regulatory and funding context limit MG modularity, leading poor customers to eventually bear the cost of ill-suited sizing.