We address the problem of compiling bathymetric data sets with heterogeneous coverage and a range of data measurement accuracies. To generate a regularly spaced grid, we are obliged to interpolate sparse data; our objective here is to augment this product with an estimate of confidence in the interpolated bathymetry based on our knowledge of the component of random error in the bathymetric source data. Using a direct simulation Monte Carlo method, we utilize data from the International Bathymetric Chart of the Arctic Ocean database to develop a suitable methodology for assessment of the standard deviations of depths in the interpolated grid. Our assessment of random errors in each data set are heuristic but realistic and are based on available metadata from the data providers. We show that a confidence grid can be built using this method and that this product can be used to assess reliability of the final compilation. The methodology as developed here is applied to bathymetric data but is equally applicable to other interpolated data sets, such as gravity and magnetic data.