Metrics are measurements. A software development project is an experiment that is based on a number of assumptions theories and estimates. To confirm these measurements are taken for all significant processes and then analyzed. The results of the analysis are used to make decisions. For example if the assumed code productivity for a project was 800 lines per person month and the actual measured average is 300 it is probably not a good idea to use 800 to estimate the next project. Determining this average is a metrics analysis task. The effort data might show that a lot of effort was spent fixing problems during system test. The data might indicate that this could be reduced by spending more effort on code inspection. Confirming this relationship is a metrics analysis task.
image credit: typepad.com