Thanks to its empirical success, the gravity approach is widely used to explain trade patterns between countries. In this article we question the simple application of this approach to product/sector-level trade on two grounds. First, we demonstrate that the traditional Armington version of gravity must be altered to properly account for the fact that sector expenditures are not strictly equal to sector productions because some trade costs are incurred outside the sector of interest. Secondly, we show empirically that collecting/using good data on sector-level trade and expenditure is extremely crucial for the quality of econometric estimations. Above all one should strictly adhere to the requirements of the theoretical frameworks in order to obtain unbiased and accurate estimates of the different parameters.