Massive machine type communications (mMTC) and mission-critical MTC are new service categories in 5G to support Internet of Things (IoT). MTC-based services, such as sensing, metering, and monitoring requires small volume of information in most cases. Since the current data transmission principle requires long codeblock to maximize the coding gain and hence novel transmission mechanism to support short packet transmission is required. In this paper, we propose a new type of data transmission scheme suitable for the ultra short packet transmission, called sparse vector coding (SVC). Key idea behind the proposed technique is to transmit the information after the sparse transformation. By mapping the information into the sparse vector and then transmitting it after the non-orthogonal random spreading, we cast the symbol detection problem into the sparse signal recovery problem in compressed sensing. Through performance analysis and simulations, we show that SVC is very effective in short packet transmissions.