Guidelines are not only sources of condensed evidence. They also contain graded recommendations that have been agreed upon in consensus procedures. Frequently there are discrepancies between the results of the critical appraisal of the evidence expressed in levels of evidence and the strength of the resulting recommendations. Unfortunately, the reasons for these discrepancies remain largely obscure and guideline users are therefore unable to comprehend this so-called 'upgrading' or 'downgrading', respectively. A consensus process consisting of several stages and a more detailed documentation could be a possible remedy. The present contribution is based on a lecture held at the 7th Annual Meeting of the DNEbM (German Network for Evidence-based Medicine) in Bochum in March 2006.