This paper explores the problem of determining the success of students in higher engineering programmes. The best way to do this is a full cohort analysis, which yields information such as throughput rates, average time to graduation and average student years per graduate. A much cruder measure is graduates/headcount, the inverse of which is similar to average student years per graduate. The paper proposes an alternative to a full cohort analysis which looks back from graduation to predict the same parameters. These analyses are compared using data for a fouryear engineering programme. The results clearly show the shortcomings of the headcount analysis, as well as the potential of the looking back analysis, especially where large sets of data are concerned. The paper also indicates the value of determining the number of student years it take to generate a single graduate, in terms of the funding which needs to be provided.