We provide new evidence on the drivers of the pass-through of exchange rate movements into consumer prices across four decades and close to a hundred countries, combining econometrics and random forests. Random forests are particularly useful for modelling highly non-linear relationships, as well as for identifying the relative importance of the different theoretical factors that can affect the degree of pass-through. We find that the size of the economy, which tends to be related to the extent of pricing-to-market, and the level of inflation emerge as the factors most strongly associated with exchange rate pass-through, followed by product homogeneity and the volatility of the exchange rate. As we show, several of these covariates display a non-linear relation with exchange rate pass-throughs. We also document important implications of macroeconomic policy regimes and outcomes, including those related to fiscal policy, for exchange rate pass-through.