Geo-fencing is a location based service that allows sending of messages to users who enter/exit a specified geographical area, known as a geo-fence. Today, it has become one of the popular location based mobile marketing strategies. Currently, the process of geo-fencing is manual, i.e. a retailer has to specify the location and the radius of area around it to setup the geo-fences. In addition, this process does not take into account the user preference towards the targeted product/service and can compromise his/her experience by unnecessarily. We attempt to solve the problem by presenting a systematic approach to define smarter geo-fences, which incorporates user preferences. The uniqueness of our definition of user preference is that it not only depends on the product/service in question but also varies with location, which has not been addressed in the literature. We term such a preference as location sensitive product affinity (LSPA) and propose a formulation to estimate this affinity. Our geo-fence creation process then leverages this affinity distribution, over an area, to create smarter geo-fences by selecting contiguous group of locations where the affinity is high. We validate our approach on a real e-commerce data and show that our approach fares well against standard industry practice as well as one of the popular collaborative filtering techniques. We thus, prove the effectiveness of our approach in designing smart geo-fences and believe it can help improve the user experience by targeting them where they are most likely to have a higher preference towards the product. Our proposed approach can have a profound impact in the field of location based mobile marketing and pave the way for further research in this area.