Arguably one of the most valuable tools for investigating pupil behaviour in an educational environment is systematic classroom observation. Classroom observation is often cited as having the potential to enable research of the learning process in action. Low inference classroom observation instruments are designed to record a sequence of data points consisting of successive measurements made over a time interval, making them particularly appropriate as measurement tools of the latent learning process. This study aimed to demonstrate that the frequently used method of analyzing this inherently temporal data proportionally is failing to consider variability between individual pupils that is identifiable in the original time series. Using a combination of real and simulated data, substantially more unique patterns of behaviour were found with time series analyses of the same data than with proportional analyses. It was also found that proportional analyses explained very little variability in the time series data of the same pupils. Modifying the number of samples and categories in an observation schedule did not vastly improve the relationship between proportional and time series analyses. We argue that if the methodological focus of classroom observation is on examining differences between pupils, then methods that maintain temporal aspects are preferable.