In recent years, analysis of flowback data has provided an early opportunity to evaluate key fracture parameters (e.g. fracture permeability and fracture volume) in unconventional reservoirs. A combination of diagnostic plots, straight-line methods and simulation model history-matching techniques have been used for this purpose. For many shale gas wells, immediate gas and water production occurs on flowback. In this paper, a novel method is developed and demonstrated for analysis of water data during two-phase (gas + water) flowback and early-time production of shale gas wells. In previous studies, it has been demonstrated that, during the late-time water flow period, the water phase exhibits boundary-dominated flow (BDF) characteristics due to the limited volume of water in fractures or surrounding matrix. The gas phase exhibits transient linear flow (TLF) characteristics within matrix, which is one of the first flow-regimes typically observed after flowback during online production. By introducing a new pseudotime with a modified total compressibility, the water flow equation during two-phase (gas + water) flow period is linearized. Accordingly, a modified flowing material balance (FMB) equation is developed for analyzing water boundary-dominated flow which considers water relativity permeability and the saturation- and pressure dependent total compressibility. Thus, a novel method is developed by combination of the water FMB equation and an iteration technique, to estimate the facture half-length and fracture permeability during flowback of shale gas wells. The approach is validated using several numerically-simulated cases. Results show that the accuracy of the new method is influenced by the pressure and saturation gradient in the fracture. With an increase in fracture permeability, the calculated fracture parameters become very close to the expected values. The calculated fracture half-length starts to deviate from the expected values with an increase in fracture half-length. However, the derived fracture permeability values are in excellent agreement for all cases studied. The matrix permeability has little effect on the calculated fracture parameters (half-length and permeability). Although there are some deviations between the interpreted and excepted values, both the values generally agree within 5%. Finally, practical application of the method is demonstrated using a Marcellus Shale well in northeast US north-east and a Changning Shale well in southwest China. The new techniques demonstrate the importance of accounting for phase relative permeability and correcting total compressibility, even when the gas phase dominates the flowback period.