This paper describes the development of an online database which allows users to assess the accuracy of ABARES agricultural market forecasts for around 100 variables over nearly two decades. Accuracy underpins the broader quality dimensions of ABARES forecasts such as institutional alignment and value-in-use to end users. The accuracy of ABARES forecasts generally improves as the lead time between forecast and outcome reduces, and production forecasts are slightly more accurate than corresponding price or export forecasts. Overall results show that ABARES forecasts are generally unbiased, but that bias can be a transient issue in markets undergoing structural change. The ability to analyse accuracy at low cost is a foundational step towards future research into the value of ABARES forecasts for supporting decision making.