This paper draws on recent research into these issues in the commercial research sector; specifically, in the measurement of television viewing. Different analyses for detecting measurement errors associated with time-in-sample are described. Panel samples offer unique atrractions for the measurement of change. Estimates of net change have low sampling variation, while gross change and changes at the individual respondent level may also be measured. Their drawbacks, at least for panels of households or individuals, include low response rates (compared with single-interview surveys), sample attrition and and the possibility of conditioning. In such panels, attrition rates vary between demographic subgroups and this can give rise to sample imbalance and bias in estimation. Conditioning also gives rise to bias. Even in the absence of measurement error, differential rates of attrition demand an active replacement sampling policy if a representative panel is to be maintained. Sample rotation represents one method. An optimal method for introducing rotation is described. However, rotation incurs higher cost and sample turnover rates than a selective discard policy. As a result rotation has seen limited application to date.