Here we present an operational method to improve accuracy and information content of ground-based measurements of stratospheric NO2. The motive is to improve the investigation of trends in NO2, and is important because the current trend in NO2 appears to contradict the trend in its source, suggesting that the stratospheric circulation has changed. To do so, a new software package for retrieving NO2 vertical profiles from slant columns measured by zenith-sky spectrometers has been created. It uses a Rodgers optimal linear inverse method coupled with a radiative transfer model for calculations of transfer functions between profiles and columns, and a chemical box model for taking into account the NO2 variations during twilight and during the day. Each model has parameters that vary according to season and location. Forerunners of each model have been previously validated. The scheme maps random errors in the measurements and systematic errors in the models and their parameters on to the retrieved profiles. Initialisation for models is derived from well-established climatologies. The software has been tested by comparing retrieved profiles to simultaneous balloon-borne profiles at mid-latitudes in spring.