The problem of designing the state estimation structure for staged process is addressed within a geometric estimation (GE) framework, with the class of binary columns as case study. The structure consists of the complete or reduced model, the number of measurements and their locations, and the innovation mechanism. The structure is chosen to execute a specific estimation task with a suitable compromise between state reconstruction speed, robustness, and number of ODEs. First, structure selection guidelines are derived from the examination of the staged system characteristics followed by the application of detectability measures, The application of the guidelines yields candidate structures. Conclusive structure results are obtained through (geometric and simplified EKF) estimator functioning assessment. The proposed approach is illustrated and tested with an experimental binary column. With less ODEs and a more systematic tuning, the behaviors of the geometric and simplified EKF estimators match the one of the conventional EKF. (C) 2012 Elsevier Ltd. All rights reserved.