# Seed Selection and Social Coupon Allocation for Redemption Maximization in Online Social Networks

## TL;DR

This paper addresses the joint problem of seed selection and social coupon allocation in online social networks to maximize redemption rates, proposing a new formulation, algorithms, and theoretical guarantees validated by experiments.

## Contribution

It introduces the S3CRM problem, proves its hardness, and develops an effective algorithm with performance guarantees for optimizing social coupon campaigns.

## Key findings

- The proposed algorithm outperforms baseline methods.
- The notion of marginal redemption effectively guides investment decisions.
- Guaranteed paths help balance seed and coupon investments.

## Abstract

Online social networks have become the medium for efficient viral marketing exploiting social influence in information diffusion. However, the emerging application Social Coupon (SC) incorporating social referral into coupons cannot be efficiently solved by previous researches which do not take into account the effect of SC allocation. The number of allocated SCs restricts the number of influenced friends for each user. In the paper, we investigate not only the seed selection problem but also the effect of SC allocation for optimizing the redemption rate which represents the efficiency of SC allocation. Accordingly, we formulate a problem named Seed Selection and SC allocation for Redemption Maximization (S3CRM) and prove the hardness of S3CRM. We design an effective algorithm with a performance guarantee, called Seed Selection and Social Coupon allocation algorithm. For S3CRM, we introduce the notion of marginal redemption to evaluate the efficiency of investment in seeds and SCs. Moreover, for a balanced investment, we develop a new graph structure called guaranteed path, to explore the opportunity to optimize the redemption rate. Finally, we perform a comprehensive evaluation on our proposed algorithm with various baselines. The results validate our ideas and show the effectiveness of the proposed algorithm over baselines.