Online access panels are becoming of paramount importance in marketing research, and constitute a great asset for market research firms. In this paper, we show that traditional models fail to quantify the true “cost” of an electronic solicitation, and we demonstrate that each additional solicitation not only decreases the likelihood of future participation, but might even increase the drop-out rate (a mostly unobserved phenomenon that has a dramatic impact on the lifetime value of an e-panelist). Our model estimates the likelihood that an e-panelist will respond positively to a new solicitation, as a function of his past behavior and how many times he has been solicited so far, and integrates a latent “wear out” effect. We fit the model on a sample of more than 700,000 e-mail solicitations sent over a period of 3 years, and demonstrate that each additional solicitation contributes to a long-lasting wear-out effect; the unobserved drop-out rate can reach up to 10% at each additional solicitation.
DE BRUYN, A. and MUKHERJEE, P. (2013). Modeling Unobserved Drop-out Rate to Optimize e-Panelist Lifetime Value. In: INFORMS Marketing Science Conference.