Determining the exchange rate pass-through on inflation is a necessity for central banks as well as for firms and households. This is an apparently easy and intuitive task, but it faces high complexity and uncertainty. This paper examines the short and long-term impact of an exchange rate shock on inflation along the distribution chain in the presence of endogeneity, nonlinearity and asymmetry. The econometric model is a smooth transition autoregressive vector estimated by Bayesian methods. This incorporates a model of pricing and the endogenous nature of the exchange rate pass-through (PT). The paper uses monthly data from Colombia for the period 2002 to 2015. The main findings are that PT is incomplete (as predicted and found by most of the recent literature) endogenous and then changes over time, nonlinear and asymmetric in the short and long terms to the state of the economy (i.e., PT is nonlinear state-dependent) and to exchange rate shocks. Historically, findings have showed that the accumulated PT on inflation of import prices rises from 20% in the first month of the exchange rate shock to a maximum of around 66% in the first year. The equivalent figures on the inflation of producer goods go from 13% to 52%; on the inflation of imported consumer goods from 6% to 48%, and on the CPI inflation from 4% to 30%. At four years, the respective figures for accumulated PT are 98%, 84%, 94% and 80%, but uncertainty about these estimates increases rapidly over time.