Integrating Social Network Effects in Product Design and Diffusion
MetadataShow full item record
Connectivities among people are amplified with recent advancements in internet technology increasing the number of communication channels. Information spread over these networks strengthen the social influence among individuals and affect their purchasing decisions. In this thesis, we study three problems in the product design and diffusion context by integrating such social network effects where influence takes place over neighborhood relationship ties among the users of the product. We consider the setting where peer influence plays a significant role in a consumer's product choice or there is a tangible benefit from using the same product as the rest of one's social network. Building upon the well-known Share-of-Choice problem, we model an influence structure and define the Share-of-Choice problem with Network Effects. It is an NP-Hard combinatorial optimization problem which we solve using a Genetic Algorithm. Using simulated data we show that ignoring social network effects in the design phase of a product results in a significantly lower market share for a product. Our genetic algorithm obtains near-optimal solutions and is very robust in terms of its running time, scalability, and ability to adapt to additional constraints/variations of the model. In this setting, we introduce a product diffusion problem, the Least Cost Influence Problem, which increases the market share of a product by intervening the natural diffusion of it over the social network. This intervention is in the form of incentive supply to a group of people in a least costly way while maximizing the spread of the product. We generalize the Least Cost Influence Problem by moving away from the marketing setting and by treating the previous product as any piece of "information" that can spread over a social network by adoption. We show that this problem is polynomially solvable over tree networks under some conditions. We provide a Dynamic Programming algorithm to solve this problem and show that it can be interpreted as a greedy algorithm that gives incentives starting with the people that are least influenced by their neighbors, albeit the definition of susceptibility to influence from neighbors is updated throughout the algorithm. We introduce a two dimensional influence model and extend our modeling and solution methods for the product line design problem which involves designing multiple products within the same product line with the objective of appealing to the heterogeneous structure of the market. The first dimension of influence is the affection of individuals from using the same product, and the second dimension is the influence of using a similar product from the same product line which has a lower intensity of influence. We reexamine the Least Cost Influence Problem in the product line setting.