This paper presents a numerical investigation of factors that influence the design of precise measurement networks based on the random error reduction attributes of data reconciliation. The cited factors provide a basis for developing design heuristics that target specified streams for maximal variance reduction. In this investigation, distinct network structures were generated from a case study flowsheet using a strategy that conserves the input and output stream configuration across all generated candidates. The factors were assessed by determining the extent of variance reduction experienced after data reconciliation. Input and output flows were targeted for optimisation owing to their custodial importance in high precision measurement applications such as metal accounting. The results confirm the dependence of variance reduction on measured variance as concluded in previous studies. However, the ratio of stream to parent-node variance emerges as a better predictor of variance reduction for individual streams, particularly in complex network structures. (c) 2012 Elsevier Ltd. All rights reserved.