A huge increase in the amount of data consumed by smartphone users is becoming a serious problem for mobile operators. In three years, mobile data traffic in the AT&T's network rose 5000%. The US operators invest 50 billion dollars in their data networks every year and the technology upgrades and innovation still fail to follow the demand. In this paper we design two algorithms for delay tolerant offloading of bulky, socially recommended content from 3G networks. The first one, called "MixZones", uses opportunistic, ad hoc transfers between users, assisted by predictions made by the network operator. The second one, called "HotZones", exploits delay tolerance and tries to download contents when users are close to Wi-Fi access points; it is also assisted by predictions made by the operator. We evaluate both algorithms using a large data set, obtained from a major mobile operator and a realistic application similar to Apple's Ping music social network. The metrics address amount of offloading, delay and mobile energy efficiency. We find that both solutions succeed in offloading a significant amount of traffic, with positive impact on user battery lifetime. Surprisingly, we also find that all the benefit obtained from the operator with the MixZones algorithm (i.e with ad hoc exchanges between users) can be achieved with the HotZones algorithm and a small investment in Wi-Fi access points. Note that the latter is probably considerably less complex to deploy than the former.