In an analysis of variable rate technology (VRT) it is not the temporal-spatial variability in yield per se that matters, but rather the temporal-spatial variability in the response to managed inputs. There are two critical questions for researchers: 1) is the ‘true' optimum input rate applied via VRT more profitable than a uniform rate?, 2) how accurately can we estimate this ‘true' optimum rate given n years of experimentation?. We address the second by analyzing a 21-year experiment of corn response to N. By generating 1000 potential experiments of 2 to 15 years, we computed the uncertainty about the ex-ante optimum N rate. For 2 years of experimentation, there is nearly a 50% probability of committing absolute errors of 40 + kg/ha for the ex-ante N rate, whereas 7 years reduce it to 10%. The economic loss associated was small due to the ‘flatness' of the profit response function around the optimum N. For 2 years of experimentation, there is a probability of about 25% of losing 10 + U$S /ha or more due to errors in estimation the optimum N, but for durations of only 4 years the probability of losing this amount is less than 9%. These amounts are comparable to the profits reported for VRT. Early (before V6) and late-season weather variables were included in a nonlinear mixed model to explain inter-annual variations in corn response to N and produced a pseudo R2 of 0.78. Utilizing this model and long-term weather records, we simulated 109 years of corn response to N. Inter-annual weather-based optimum N rates ranged between 120 and 190 kg/ha versus a ‘flat' non-weather based ex-ante optimum of 171 kg/ha. The adjustment can thus have significant impacts within individual years, although the long-term economic gain was less than 1 U$S/ha.