Not logged in.
Quick Search - Contribution
Contribution Details
Type | Conference Presentation |
Scope | Discipline-based scholarship |
Title | Will it spread? Quantifying the predictability of new product diffusion in social networks |
Organization Unit | |
Authors |
|
Presentation Type | speech |
Item Subtype | Original Work |
Refereed | Yes |
Status | Published electronically before print/final form (Epub ahead of print) |
Language |
|
Event Title | Informs Marketing Science |
Event Type | conference |
Event Location | Duke University, USA (online) |
Event Start Date | June 11 - 2020 |
Event End Date | June 13 - 2020 |
Abstract Text | The opportunity to capitalize on social contagion has led many firms to invest significant resources in designing viral products and identifying the best seeding strategies. While extensive literature has been devoted to addressing both topics, incorporating this knowledge to predict and engineer product virality remains a difficult task. In this article, we examine whether the diffusion of a new product can be predicted based on individual, product and social network characteristics. To this end, we integrate a lab experiment with an agent-based model of product diffusion, and validate our results on empirical data. In the lab experiment, we use a conjoint design to measure the individual susceptibility to social influence from observed product choices. We show that susceptibility is dependent on the interplay between product and individual characteristics. We use the experimental results to calibrate an agent-based model of new product diffusion in a social network. We quantify the success predictability of different products, the potential outcome and risk associated with seeding strategies, and the role played by product and network characteristics on cascade size. Furthermore, we propose a method to construct an optimal portfolio of seed nodes with an ordinary number of contacts, and show that it outperforms seeding high degree nodes. We validate our results on susceptibility inference and diffusion predictability in an empirical study of an online food community (1M users) where we observe the diffusion of over 75’000 user-generated recipes over 10 years. Overall, our findings shed light on the drivers of social contagion, establish a link between micro-level observations and macro-level outcomes, and provide insight into designing more effective viral marketing campaigns. |
Export | BibTeX |