A key question in molecular genetics is why severe mutations often do not result in a detectably abnormal phenotype. This robustness was partially ascribed to redundant paralogs that may provide backup for one another in case of mutation. Mining mutant viability and mRNA expression data in Saccharomyces cerevisiae, we found that backup was provided predominantly by paralogs that are expressed dissimilarly in most growth conditions. We considered that this apparent inconsistency might be resolved by a transcriptional reprogramming mechanism that allows the intact paralog to rescue the organism upon mutation of its counterpart. We found that in wild-type cells, partial coregulation across growth conditions predicted the ability of paralogs to alter their transcription patterns and to provide backup for one another. Notably, the sets of regulatory motifs that controlled the paralogs with the most efficient backup activity deliberately overlapped only partially; paralogs with highly similar or dissimilar sets of motifs had suboptimal backup activity. Such an arrangement of partially shared regulatory motifs reconciles the differential expression of paralogs with their ability to back each other up.