We show that it is possible to describe the effective Pomeron intercept, determined from the HERA deep inelastic scattering data at small values of Bjorken x, by using next-to-leading order Balitsky-Fadin-Kuraev-Lipatov evolution together with collinear improvements. To obtain a good description over the whole range of Q(2), we use a non-Abelian physical renormalization scheme with the Brodsky-Lepage-Mackenzie optimal scale, combined with a parametrization of the running coupling in the infrared region.