The traffic in current wireless networks exhibits large variations in uplink (UL) and downlink (DL), which brings huge challenges to network operators in efficiently allocating radio resources. Dynamic time-division duplex (TDD) is considered as a promising scheme to flexibly adjust resource allocation based on UL and DL traffic demands, known as traffic adaptation. In this work, we study how traffic adaptation decreases cell service time and improves energy efficiency (EE) in small cell networks operating dynamic TDD. According to different UL and DL traffic parameters, we classify small cells into K ≥ 1 types, and accommodate the UL/DL configuration for each type of small cells with the objective to minimize the cell service time and maximize the network EE. In comparison with semi-static TDD scheme, dynamic TDD is shown to achieve larger service time gain as the traffic asymmetry between small cells increases. In summary, the proposed analytical framework allows us to elucidate the benefit of traffic adaptation to service time and EE in future dense networks with dynamic TDD.