Why do I receive different performance data with different speed tests?

Every speed test provider has its own measuring method and data rate calculation. Unfortunately, there is currently no consistent description of a performance test. cnlab periodically compares the public results of various speed test providers with its own measurements in order to assess comparability of the performance tests.

How many measurements are performed and how statistically indicative are the results?

Between 10,000 and 20,000 tests are performed every day by 3,000 to 4,000 different users. Upon announcements in the media or if the offers change, up to 10x more measurements can be recorded. For statistical evaluations, the measurements are summarized per user or per internet connection. The results of testers who performed a great many tests will therefore not be weighted more than the results of testers who only performed a few tests. In the statistical evaluations, it is always observed that at least 100 values are available for the effect observed.

The measurements lead to rather pessimistic values for two reasons: On the one hand, users test mostly when they feel that there are any performance issues; on the other hand, they test mostly during "high-load periods". Based on information from Cablecom, it could be verified in 2007 that about 59 measurements per 100 customers were performed in regions known for a bad performance, while only 27 measurements per 100 customers are observed in areas with a good performance. It becomes clear that grids with performance issues will lead to more customer measurements.