This article addresses the issue of favoritism at the design stage of a complex procurement auction. A community of citizens procures a project but lacks the ability to translate its preferences into operational technical specifications. This task is delegated to a public officer who may collude with one of the firms in exchange of a bribe. We investigate a simple accountability mechanism that requires justifying one aspect of the technical decision determined by the alerts of competitors (alert-based accountability [ABA]). We find that relying on competitors enables the community to deter favoritism significantly more easily than random challenges. The penalty needed to fully deter corruption is independent of the complexity of the project. It depends on the degree of differentiation within the industry. In an illustrative example, we study the patterns of favoritism when corruption occurs under ABA and compare them with the patterns in the random challenge mechanism.