Optimising the sowing date of canola (Brassica napus L.) in specific environments is an important determinant of yield worldwide. In eastern Australia, late April to early May has traditionally been considered the optimum sowing window for spring canola, with significant reduction in yield and oil in later sown crops. Recent and projected changes in climate, new vigorous hybrids, and improved fallow management and seeding equipment have stimulated a re-evaluation of early-April sowing to capture physiological advantages of greater biomass production and earlier flowering under contemporary conditions. Early–mid-April sowing generated the highest or equal highest yield and oil content in eight of nine field experiments conducted from 2002 to 2012 in south-eastern Australia. Declines in seed yield (–6.0% to –6.5%), oil content (–0.5% to –1.5%) and water-use efficiency (–3.8% to –5.5%) per week delay in sowing after early April reflected levels reported in previous studies with sowings from late April. Interactions with cultivar phenology were evident at some sites depending on seasonal conditions. There was no consistent difference in performance between hybrid and non-hybrid cultivars at the earliest sowing dates. Despite low temperatures thought to damage early pods at some sites (<−2°C), frost damage did not significantly compromise the yield of the early-sown crops, presumably because of greater impact of heat and water-stress in the later sown crops. A validated APSIM-Canola simulation study using 50 years of weather data at selected sites predicted highest potential yields from early-April sowing. However, the application of a frost-heat sensitivity index to account for impacts of temperature stress during the reproductive phase predicted lower yields and higher yield variability from early-April sowing. The frost–heat-limited yields predicted optimum sowing times of mid-April at southern sites, and late April to early May at the northern sites with lower median yield and higher yield variability in crops sown in early April. The experimental and simulation data are potentially compatible given that the experiments occurred during the decade of the Millennium drought in south-eastern Australia (2002–10), with dry and hot spring conditions favouring earlier sowing. However, the study reveals the need for more accurate and validated prediction of the frost and heat impacts on field-grown canola if simulation models are to provide more accurate prediction of attainable yield as new combinations of cultivar and sowing dates are explored.