This tool enables comparing performance of a given kubemark test against that of the corresponding real cluster test. The comparison is based on the API request latencies and pod startup latency metrics. And it is based on aggregate results of the tests over multiple runs. While using this tool for kubemark performance is the prime usecase, this can in general be used for comparing any two tests, making most sense when the tests run in similar environments (i.e. with similar no. of nodes, e2e tests, etc).
For further details, here's the design doc for the tool.