A new framework is presented for analysing the causes of model sea ice biases, demonstrated with the CMIP5 model HadGEM2-ES. Arctic sea ice extent has decreased over recent decades and has reached historic minima in late summer in recent years. Climate models project an ice-free Arctic in late summer during the 21st century, with wide-ranging implications for global climate and geopolitics. However, substantial spread remains in climate model projections of the rate of sea ice decline, with drivers poorly understood. In the framework described, the sea ice volume is treated as a consequence of the integrated surface energy balance. A system of simple models allows specific portions of the surface flux anomaly to be attributed to individual processes by calculating for each process an induced surface flux anomaly. The method allows detailed quantification of the role played by the surface albedo and ice thickness-growth feedbacks in causing anomalous sea ice melt and growth. It shows biases in the HadGEM2-ES sea ice volume simulation to be due to a bias in spring surface melt onset date, partly countered by a counteracting bias in winter downwelling longwave radiation. The framework is applicable in principle to any model and has the potential to greatly improve understanding of the reasons for ensemble spread in modelled sea ice state.