Temporal aggregation of the traffic stream characteristics has a significant influence on the performance of any mathematical model derived thereafter. For a heterogeneous traffic stream, the selection of the data aggregation interval is further complicated due to the possibility of different vehicle mixes in a series of composition data collected over a fixed aggregation period. The present study critically investigates and models the variation of the vehicle composition in the time series data collected over different aggregation periods. Traffic composition for a particular aggregation interval was represented using the coefficient of variation of the traffic composition corresponding to that aggregation interval. Findings from this study reveal that the traffic composition variation with the data aggregation interval could be modeled as a rational function. From the modeled variation, the minimum aggregation interval corresponding to the consistent traffic composition could be estimated. The present study suggests that the minimum aggregation interval could be approximated corresponding to the aggregation interval at which the rate of change of the coefficient of variation approaches zero from the direction of the lower aggregation period. The variation of the traffic composition within this aggregation interval would be minimal, hence safeguard the data from a wide scattering caused by the vehicle mix. © 2020 The Authors. Published by Elsevier B.V.