Kinetic data measured at low temperatures are very sparse; for instance, in state-of-the-art photo-chemical models of Titan atmosphere, less than 10% of the rates have been measured in the relevant temperature range (100-200 K). A consequence is that photochemical models rely mostly on low-T extrapolations by Arrhenius-type laws. There is more and more evidence that this is often inappropriate, and low T extrapolations are hindered by very high uncertainty. The predictions of models based on those extrapolations are expected to be very inaccurate. A dialogue between modellers and experimentalists is necessary to improve this situation. Considering the difficulty of low-T reaction kinetics experiments, the identification of key reactions has to be based on an optimal strategy to improve the predictivity of photochemical models. This can be achieved by global sensitivity analysis, as illustrated on Titan atmospheric chemistry. The main difficulty of this scheme is that it requires a lot of inputs, mainly the evaluation of uncertainties for extrapolated reaction rates. Although a large part has already been achieved by Hébrard et al., extension and validation requires a group of experts. A new generation of collaborative kinetic database is needed to implement efficiently this scheme. The KIDA project, initiated by V. Wakelam for astrochemistry, has been joined by planetologists with similar prospects. A result from these meetings was the reevaluation of the uncertainties associated with measured low-T reaction rates. The impact of this database update on the predicted densities of neutrals species in Titan's atmosphere is remarkable. There is a general precision improvement, but some species densities remain poorly predicted. Using these results, we generated a new list of key reactions, i.e. having a strong impact on the precision of model predictions, for which better estimation will have to be obtained, either experimentally or theoretically.