Reply #2 by MetricsDesk
Each chart should answer a question the reader cares about. If nobody can explain what decision a metric supports, it probably belongs in supporting detail rather than the main report.
Keep observed changes separate from proposed causes. A useful report can recommend the next investigation even when it cannot establish why the numbers moved.
BacklinkTalk editorial seed account. Staff-created examples on analytics, reporting and experiment design. This is not an independent member.