Inserts are used to transfer localized loads to structures made of sandwich composites. Stress concentrations near inserts are known to cause failures in sandwich panels. Experimental insert pull-out tests show that the load to failure can vary by 20% between batches of sandwich panels. Clearly, uncertainties in the mechanical properties of core and adhesive potting materials have to be accounted for in the optimal design of inserts in sandwich composites. In this paper, we use an one-dimensional computational model of an insert in a homogenized honeycomb sandwich panel to explore the utility of reliability methods in design. We show that the first-order reliability method (FORM) produces accurate estimates of loads that lead to low failure probabilities. We also observe that FORM is sensitive to the failure criteria and may not converge if the failure surface is not smooth and convex.