In recent years, the measurement-oriented Internet of Things (IoT) services has become a prominent solution in realizing digital evolution. Loss of data during measurements in IoT platforms is one of the main problems, that ultimately degrade the performance of the IoT services. Real-world measurement data has numerous dissimilar features such as amplitude resolutions, sampling rates, the number and quality of sensors deployed, data collection network, the availability and extent of ground-truth annotations. Those heterogeneities pose a significant challenge for researchers intending to use data imputations because of the required data characteristics, acceptable error, available computing resources and time restriction of processing. In short, there is a lack of widely agreed unique data imputation techniques and its useability depends upon various factor such as cause of missing and data characteristics, the nature of measurement areas and IoT applications, missing mechanism & proportion, computational complexity and prediction accuracy. We give a model to foster the selection criteria of data imputation techniques for different measurement environments, which gives a greater benefit to the IoT application and service building community.