Abstract The flux–flux plot (FFP) method can provide model-independent clues regarding the X-ray variability of active galactic nuclei. To use it properly, the bin size of the light curves should be as short as possible, provided the average counts in the light-curve bins are larger than ∼200. We apply the FFP method, performed in 2013, to the simultaneous XMM–Newton and NuSTAR observations of the Seyfert galaxy MCG–6-30-15, in the 0.3–40 keV range. The FFPs above ∼1.6 keV are well described by a straight line. This result rules out spectral slope variations and the hypothesis of absorption-driven variability. Our results are fully consistent with a power-law component varying in normalization only, with a spectral slope of ∼2, plus a variable, relativistic reflection arising from the inner accretion disc around a rotating black hole. We also detect spectral components that remain constant over ∼4.5 d (at least). At energies above ∼1.5 keV, the stable component is consistent with reflection from distant, neutral material. The constant component at low energies is consistent with a blackbody spectrum of kT BB ∼ 100 eV. The fluxes of these components are ∼10–20% of the average continuum flux (in the respective bands). They should always be included in the models that are used to fit the spectrum of the source. The FFPs below 1.6 keV are non-linear, which could be due to the variable warm absorber in this source.