Statistics South Africa (Stats SA) has been collecting labour market data with household surveys and in a fairly comparable format since 1993. These datasets have been studied and compared extensively in order to better understand the workings of the South African labour market. Many of these studies compare household surveys of different periods in order to identify trends, but the validity of such trends is conditional on the comparability of the different datasets. Besides, the naïve comparisons of the different datasets have been questioned. Other problems include inconsistencies in questionnaire design, coding errors, changes in the sampling frame, the oversampling of agricultural workers in OHS1995, the oversampling of subsistence agricultural workers in LFS2000a and LFS2000b, as well as the oversampling of informal workers in LFS2001a. Most of these issues have received attention in papers by Burger and Yu (2006), Casale, Muller and Posel (2005), and Wittenberg (2004). By drawing attention to a few of the lesser known problems, this paper aims to build on the existing literature by further stimulating debate around the strengths and weaknesses of the existing survey data, as well as considering the best ways in which to analyse the existing data. The inconsistencies that occur in the data independently of the way in which questions are asked by the interview, as well as the inconsistencies that result from the way in which the survey questions are formulated or placed in a given sequence are discussed. Where possible, adjustments that may contribute towards increased consistency in the responses are suggested. Ultimately, it is hoped that the lessons learnt from such discussions will serve to inform questionnaire design in future.