This study explains the systematic changes in the parameter estimates of the Bass model as the number of data points increases by the level of the underlying heterogeneity of the population. Using the gamma/shifted Gompertz model as the "true" model of diffusion, the study shows that when the data exhibit more right skew (more heterogeneity) than the Bass model, the Bass model estimates of p and c decrease and that of q increases as one adds data points, in contrast, when the data exhibit more left skew (less heterogeneity) than the Bass model, the Bass model estimates of p and c increase and that of q decreases as one adds data points. These results are consistent with those obtained with real data. Link to the article
BEMMAOR, A.C. and LEE, J. (2002). The Impact of Heterogeneity and Ill-Conditioning on Diffusion Model Parameter Estimates. Marketing Science, 21(2), pp. 209-220.