Index Policies for Campaign Promotion Strategies in Reward-based Crowdfunding

Wang, Chenguang and Li, Dong and Li, Baibing (2025) Index Policies for Campaign Promotion Strategies in Reward-based Crowdfunding. European Journal of Operational Research, 327 (2). pp. 515-539. ISSN 0377-2217

Text (Wang et al 2025 EJOR)
Wang_et_al_2025_EJOR.pdf - Accepted Version
Available under License Creative Commons Attribution.
Download (16MB)

Abstract

Reward-based crowdfunding plays a crucial role in fundraising for start-up entrepreneurs. Recent studies, however, have shown that the actual success rate of fundraising projects is surprisingly low across multiple crowdfunding platforms. This paper considers crowdfunding platforms’ decision-making of selecting projects to highlight on their homepage to boost the chance of success for projects, and investigates promotion strategies aiming at maximizing platforms’ revenue over a fixed period. We characterize backers’ investment decisions by a discrete choice model with a time-varying coefficient of herding effect, and formulate the problem as a stochastic dynamic program, which is however computationally intractable. To address this issue, we follow the Whittle’s Restless Bandit approach to decompose the problem into a collection of single-project problems and prove indexability for each project under some mild conditions. We show that the index values of the proposed index policy can be directly derived from the value-to-go of each project under the non-promotion policy, which is calculated recursively offline with a linear-time complexity. Moreover, to further enhance the scalability we develop a closed-form approximation to calculate the index values online. To the best of our knowledge, this work is the first in the literature to develop index policies for campaign promotions in reward-based crowdfunding. It is also the first attempt to provide indexability analysis of bi-dimensional restless bandits coupled by not only resource but also demand. Extensive numerical experiments show that the proposed index policies outperform the other benchmark heuristics in most of the scenarios considered.