The present paper reports experimental and theoretical studies on the structure of the metastable C precipitate forming during precipitation hardening in Al–Mg–Si–Cu (6xxx) alloys. We describe the procedure of deriving an initial unit cell model based on experimental data and how this is further refined by quantitative use of nanobeam electron diffraction patterns. A reliable 3D refinement was prevented by the small precipitate thickness and its disorder/intergrowth with other phases, necessitating the development of a more theoretically based methodology for precipitate composition determination. We find that for experimental results to be acceptably reproduced in density functional theory-based calculations on bulk candidate structures, these would have to not only minimise precipitate formation enthalpy, but also reproduce the experimentally reported negligible lattice mismatch with the Al matrix along the precipitate main growth direction. We argue, through comparison of the isostructural Q′ and Q precipitate phases of Al–Mg–Si–Cu alloys, why the experimentally reported (semi-)coherency of metastable precipitates must be included as an optimisation criterion in a general theoretical analysis. The C phase structure determined has a monoclinic unit cell with space group P21/m, cell dimensions a C = 10.32 Å, b C = 4.05 Å, c C = 8.10 Å, β C = 100.9°, coherency relations with the matrix: 150Al, 001Al, 100Al and composition Mg4AlSi3+ x Cu1− x , with x 0.3. The non-integer number of atoms in the unit cell emphasises that the concept of a unit cell for this phase is weakly defined.