I present analysis of both the early and late time radio and X-ray data of the tidal disruption event Swift J1644+57. I show that the data at early times (<~5 days) necessitates separation of the radio and X-ray emission regions, either spatially or in velocity space. This leads to the suggestion of a two component jet model, in which the inner jet is initially relativistic with Lorentz factor Gamma~15, while the outer jet is trans-relativistic, with Gamma <~1.2. This model enables self-consistent interpretation of the late time radio data, both in terms of peak frequency and flux, without the need for late time energy injection.