Background: Proteomic protein identification results need to be compared across laboratories and platforms, and thus a reliable method is needed to estimate false discovery rates. The target-decoy strategy is a platform-independent and thus a prime candidate for standardized reporting of data. In its current usage based on global population parameters, the method does not utilize individual peptide scores optimally. Results: Here we show that proteomic analyses largely benefit from using separate treatment of peptides matching to proteins alone or in groups based on locally estimated false discovery rates. Our implementation reduces the number of false positives and simultaneously increases the number of proteins identified. Importantly, single peptide identifications achieve defined confidence and the sequence coverage of proteins is optimized. As a result, we improve the number of proteins identified in a human serum analysis by 58% without compromising identification confidence. Conclusion: We show that proteins can reliably be identified with a single peptide and the sequence coverage for multi-peptide proteins can be increased when using an improved estimation of false discovery rates.