Scientific surveys are expensive and often not feasible for artisanal and small-scale fisheries. Biological samples from the commercial catch are the principal source of information in monitoring these fisheries. A pilot sampling scheme of commercial craft conducted during a four-month period on the artisanal coastal shrimp trawl fishery in Sri Lanka revealed that the length compositions of two short-lived shrimp species (Metapenaeus dobsoni and Parapenaeopsis coromandelica), derived from commercial catches, showed high sampling variance and thus resulted in considerable noise in the length-frequency distributions. An analysis of variance components revealed that the total variance was mainly governed by the variance due to the number of samples. Hence, a more effective survey design was to collect samples from as many fishing craft as possible rather than increasing the sample size. The observed total variance of 0.4 based on measuring 43 109 individuals of M. dobsoni from 4 boats per sampling day (∼300 individuals/sample) in the pilot sampling was revised with the new sampling scheme, which reduced the variance to 0.2 based on measuring only 28 800 individuals from 6 boats per sampling day (∼150 individuals/sample). The length-frequency distributions constructed with the revised sampling scheme resulted in less noise in the monthly progression of modes.