Targets and tools for the monitoring and evaluation of educational policies and schools have changed rapidly in the last 20 years. On one hand, education literature has concluded that the gold standard for measuring the true causal impact of educational interventions is randomised controlled trials. On the other hand, although impact evaluation is the mainstream for evaluating educational programmes targeted at individual level, when programmes are intended for organisations it also becomes relevant for benchmarking to measure educational efficiency and total factor productivity changes of schools by means of production frontiers. Surprisingly, to date in the economics of education, both fields of research; impact evaluation and production frontiers, run as parallel lines with scarce relationship between them. In this chapter we develop a theory to relate impact evaluation and production frontiers using the well-known education production function framework introducing the idea of a By-Group Malmquist index. To illustrate its potential, we run a Monte Carlo analysis simulating different educational policies, or treatments, to show how production frontiers can help to enhance the traditional impact evaluation. Our results show that successful policies for raising schools’ productivity can be hidden in the causal inference analysis if we only consider mean output differences between the treated and the control groups. In these cases, the treatment effects are better measured regarding total factor productivity changes because it allows us to measure efficiency and technology changes determined by best practices detected through production frontiers.