This paper examines the consequences of using so-called "real-time" data for business cycle analysis in Germany. Based on a novel data set covering quarterly real output data from 1980 to 2002 real-time output gaps using some popular filter methods are calculated. They differ considerably from their counterparts based on the most recent data and are, thus, not very reliable. While real-time output gaps are generally not unbiased forecasts of the final output gap series, they provide at least some information regarding the sign of the final output gap. The information content of output gaps calculated in real-time for future inflation is tested by means of an out-of-sample forecast exercise and found to be very limited. Generally, the results for simple growth rates appear to be more promising that the results for simple filters to estimate the output gap. This points to the possibility that the problematic nature of the real-time output gaps is not due to revisions of the underlying data but due to the end-of-sample problem that occurs in filtering recent data. All in all, the results support previous findings regarding other countries that revisions of data and output gap estimates can seriously distort business cycle analysis and, thus, research and policy decisions.