Intelligent reflecting surfaces (IRS) have become the subject of many current research efforts, as the ongoing wireless spectrum crunch has made the need to open higher frequency bands a priority. IRS are one of the alternatives proposed to overcome the problem of line-of-sight blocking in very high frequency wireless scenarios. The current state-of-the-art shows the difficulty of implementing practical IRS designs able to redirect large signal bandwidths, prone to the so-called beam split (BS) dispersion effect. In this work, we propose two highly efficient configuration techniques, adapted to ultrawideband downlink scenarios, based on localized optimization over the IRS surface. Such techniques exploit the BS effect while taking into account for the shape of the transmitted signal spectrum. Simulations considering different geometrical setups and different signal spectra show how the proposed techniques are able to guarantee an increased signal power at the receiver with respect to classical narrowband-based solutions or techniques that perform a global optimization over the entire IRS surface.