We study the mobile coupons delivery problem (MCDP) in a push-based location based advertisement where an advertiser proactively sends mobile coupons to prospective customers on behalf of stores based on the customers' location and preferences. MCDP provides a new capability called postponable selection that enables an advertiser to better capitalize on the plethora of customer information provided by mobile phone users through network service providers. Postponable selection allows a customer to be reconsidered for selection for coupon delivery (i.e. postponed) as more information becomes available. We formulate MCDP as a new problem category referred to as the postponable online multi-constraint knapsack. We propose a single threshold-type algorithm with different design options and conduct extensive computational experiments to discuss the effectiveness of our algorithm as well as the benefit of postponable selection. Our experimental results show that the proposed algorithm outperforms the First-Come-First-Serve heuristic by 46% and postponable selection additionally improves performance by 4% on average.