The problem of uncertainties in input-output coefficients is examined, using the uncertainty in estimating the fertilizer use efficiency as an illustration. An example of uncertainty due to lack of knowledge on processes involved is the use of different approaches for estimating fertilizer use efficiency in two land use optimization studies. A further problem is uncertainty due to lack of data; this is illustrated with an example from the Atlantic Zone of Costa Rica. Very few data are available to determine fertilizer use efficiency and data from regions with similar soil and climate type are not available. Data from non-similar regions may not give a correct assessment of the possibilities in the region. Different concepts and sources of information result in different estimates of coefficients, which might in turn greatly influence the results of the linear programming model. It is therefore concluded that, rather than using one fixed value for a particular input-output coefficient, the effect of uncertainty in coefficients on the final results of the model should be examined.