Western blue groper populations are at high risk of being overfished in Western Australia. At present, there is very little fisheries independent population data for this species. It is critical then, that population surveys are conducted to fill this gap using a method that is non-extractive, cost-effective and efficient. Stereo- baited remote underwater video systems are one such method, and are comparable to other fisheries dependent techniques. However, when rare fauna are the species of interest, such as the blue groper, they are often not encountered frequently enough to obtain robust estimates of abundance. This study aims to improve blue groper abundance surveys in two ways; (1) test attractiveness of alternative bait types, and, (2) to test a novel count methodology to better sample blue groper relative abundance and length frequency distribution. Our novel estimate of abundance, MaxNSIZECLASS, categorises the individual blue gropers into size classes during the video annotation process whilst ensuring no individual fish is counted more than once. We found that our novel count metric, MaxNSIZECLASS, counted significantly more blue groper (44.9%; p <0.001) than the standard MaxN. Further, MaxNSIZECLASS detected a difference in blue groper abundance between bait types while MaxN did not. MaxNSIZECLASS also provided greater detail in size distributions of blue groper across the Recherche Archipelago than the equivalent standard method. This case study shows that the novel MaxNSIZECLASS method provides more accurate and powerful estimates of blue groper abundance than the standard MaxN, and should be explored further.