Catch at length is an important input into any stock assessment. Consequently, collecting length data from the catch is a task undertaken by all at sea observes in CCAMLR fisheries. Although analyses in the past have looked to the optimal design of the observer program, in terms of levels of coverage of vessels and hauls (e.g., Agnew et al., 2009; 2010), less attention have been focussed on how many krill should be measured by observers should measure (however see Thannasekos et al. 2012). We used C1 effort, catch and observer data from Subarea 48.1, collected between 2010 and 2015, to characterise how many krill are measured by observers, and for how many hauls. We then simulate the impact of different haul-wise sample sizes on the ability to estimate mean length in a sample per SSMU × month combination (effective sample size). The median number of krill measured per haul was around 200 (range 0-652). However haul-wise sample sizes of down to 50 did not substantially reduce the effective sample size, whereas increasing the number of hauls sampled did substantially increase the effective sample size. Therefore, we recommend that observers collect smaller samples (50) at the haul level, over a greater number of hauls to allow better estimates of catch at length.