Explain a metric move without inventing a story

{{Metric}} moved from {{before}} to {{after}} between {{dates}}. Context: {{what you know — launches, campaigns, seasonality, outages}}. Data available: {{list dimensions you can segment by}}.

Work it properly:
1. Decompose the metric into its arithmetic parts and say which part moved. Show the arithmetic.
2. Mix versus rate: did behaviour change, or did the population change? Specify the segment cut that separates them.
3. List candidate causes, and for each: the segment pattern we'd see if it were true, and the pattern that would rule it out.
4. Rank by consistency with what we already observe.
5. Name the instrumentation and tracking explanations — a definition change, a broken event, a bot wave — before the behavioural ones.
6. State the single query I should run next, and what result would change your ranking.

Do not give me a narrative. Give me discriminating tests.