Understanding statistical double-averaging Both Observer Reporting Server and Observer APA- and ATA-based reports use per-server, per-time interval averaging. This means that the average response time for multiple servers is calculated based on the sum of average response time for each server divided by the number of servers for each time interval. The calculation is done this way on purpose. It allows focusing on determining the average response time for all servers, instead of the busiest server weighing the response time disproportionately by either lowering or increasing the average response time based on number of observed connections. When data is collected in Observer with higher granularity than aggregated in Observer Reporting Server, the numbers for average response time may differ from what is seen in the internal Observer Network Trending Viewer tool (and Observer Web Reports that use the same trending data). This is due to double averaging, and even though the numbers differ, they show valid statistical information based on the context. Ultimately, Observer Reporting Server results are most useful for long-term trending and averages are based on this goal, while Observer can provide higher granularity suitable for further investigation. If the Observer collection interval is equal to the Observer Reporting Server data transfer schedule, the results would be identical. Alternatively, if traffic is represented in every minute of collection for both servers, even with differing Observer and ORS intervals, the results would still be very close to each other.
