Thermal conductivity enhancement of nanofluids primarily depends upon the factors affecting their heat transport mechanisms. More insight into the phenomenon of thermal heat conduction in nanofluids has been presented in this paper. The growth of nanoclusters in the water (DI) based hanofluid containing gamma-Al2O3 (size 25-30 nm), has been investigated and studied. A comprehensive report on the suspension size distribution of the nanoparticles at different pH values, their zeta potential along with the stability ratio of suspensions, has been presented. A quantitative analysis on the thermal conductivity enhancement, along with an investigation on the role of nanoparticles present in the form of dead ends and backbone chains in an aggregate has been put forward. Their individual effect under perikinetic heat conduction conditions has been incorporated to address the thermal conductivity enhancements of Al2O3-H2O nanofluid. Moreover, the effect of the basefluid layering around the nanoparticles in an aggregate has been highlighted. The possible and the different governing static/structural models which are capable to predict the nanocluster based thermal conductivity enhancements in nanofluids, have been investigated and modified. This has been done to include the effect of liquid layering on the thermal conductivity of the water present in an aggregate compared to the bulk water present outside to a nanocluster or an aggregate. While, modifying the equations and models, the focus has been laid down to the average hydrodynamic site of the nanoparticles in a suspension than the individual particle size. The modified model gives a fairly good prediction of the thermal conductivity enhancement of the nanofluid under investigation. The experimental and theoretical results of thermal conductivity of Al2O3 nanofluid have been correlated and found to be within the accuracy level of +/- 1 to +/- 23% at 0.05% volume fraction. The error analysis shows that the developed model is enough capable to predict the thermal conductivity enhancement with an average error of +/- 2.6% compared to the predictions made from the other existing theoretical models, where the error involved is found to be varying from +/- 3 to +/- 9% at volume fraction ranging from 0.01 to 0.12%. (C) 2017 Elsevier B.V. All rights reserved.