This is the numerical indicators of how well the ML pipeline performs, either on your data, a dataset that resembles your data, or an academic benchmark. If the latter, this will also include how other competing open and closed-source tools may perform on the same benchmark.