When a KPI moves sharply, the first job is not to create a complicated dashboard. It is to establish what actually changed.
Start with the size and timing
Compare the current value with a relevant baseline. Check whether the change is sudden, gradual, seasonal or caused by a reporting cutoff.
Check the data before the business story
Look for pipeline failures, missing dates, duplicate records, changed definitions and broken joins. A dashboard change is not automatically a business change.
Segment the movement
Break the metric down by the dimensions most likely to explain it: product, customer type, geography, channel, device or time.
Separate correlation from cause
A segment that moved at the same time is a lead, not proof. Validate the mechanism before recommending action.
For a practical framework for analysis and verification, get the book on Amazon.
One response to “Root Cause Analysis for Data Analysts: How to Find Why a Metric Changed”
[…] For the broader metric-investigation process, see the root-cause analysis guide. […]