Fast electrification of the energy system and end-uses, either direct or indirect, is a cornerstone of the transition to carbon neutrality. Building an ensemble of 704 scenarios to 2050, we analyze how carbon price, technological and resource assumptions impact the arbitrage between direct and indirect electrification. Using novel explainability methods from machine learning, we attribute changes in electrification rates either to the individual or combined effects of model assumptions, or features. Total electrification is mostly driven by direct electric uses; individual effects explain up to 80% of it. Conversely, the development of indirect electrification, through hydrogen and e-fuels, is conditional on complex interaction effects between features, whose magnitude can reach up to three times the net effect in the transport sector. Moreover, we measure how interaction patterns between features reconfigure when carbon price increases. By explaining transition scenarios, our study provides sound evidence in favor of coordinated transition policy bundles.​

• ​Read the Working Paper​