Creation and improvement of tools for digital preservation is a difficult task without an established way to assess any progress in their quality. This happens due to low presence of solid evidence and a lack of accessible approaches to create such evidence. Software benchmarking, as an em- pirical method, is used in various fields to provide objective evidence about the quality of software tools. However, the digital preservation field is still missing a proper adoption of that method. This paper establishes a theory of benchmarking of tools in digital preservation as a solid method for gathering and sharing the evidence needed to achieve widespread improvements in tool quality. To this end, we discuss and synthesize literature and experience on the theory and practice of benchmarking as a method and define a conceptual framework for benchmarks in digital preservation. Four benchmarks that address different digital preservation scenarios are presented. We compare existing reports on tool evaluation and how they address the main components of benchmarking, and we discuss the question of whether the field possesses the right combination of social factors that make benchmarking a promising method at this point in time. The conclusions point to significant opportunities for collaborative benchmarks and systematic evidence sharing, but also several major challenges ahead.