Two groups, one metric, read from the observation store under your own access, over the months both groups report. Where a cohort is too small to compare safely, or the data cannot support a test, this says so rather than answering anyway.
Correlate
Metrics against metrics, across every department you can read, each averaged over the same months. Significance is corrected for looking at every pair at once. The result is association, never explanation.
0 of 8 chosen · at least two. Spearman (ranks) is the default; Pearson assumes a straight line.
Worth a look
Pick an outcome. Every other metric, from every source system, is tested against it across departments, and only what survives the correction for testing them all is listed.