Analysis Report
Workflow
Use templates/analysis-report-template.md to document every analysis.
Clarify Purpose (Section 1): What do you want to know? Why is this analysis needed? Who will use it?
Domain knowledge confirmation:
-
Confirm factual background with the user — do not assume or speculate on business context (e.g. "cost is high" vs "volume is high")
-
If analyzing a feature/initiative, confirm its exact specifications and parameters (thresholds, targets, display conditions, etc.) before designing queries
Questions to answer when comparing pre/post periods:
-
How will you ensure comparison periods are unified across all metrics?
-
Are the thresholds appropriate for business reality (weekday/weekend patterns, seasonality)?
-
Do any metrics depend on events that haven't occurred yet in the analysis period?
-
What normalization method is appropriate for volume or seasonal variation?
-
How will you account for lead time lag at period boundaries?
Discover Data (Section 2): Explore available datasets and understand schema.
-
Ask user for project/dataset context and business background
-
Use /bq-studio skill for BigQuery schema exploration
-
db/schema.rb for Rails projects
-
API docs or sample data for external services
-
Document schema and table relationships in the report
Build Query (Section 3): Use /bq-studio skill to design and execute queries.
-
Requirements and schema from Steps 1-2 provide context
-
Document query design and raw results in each subsection (SQL and results together)
Query-result consistency rules:
-
The SQL written in the report MUST be the exact SQL that was executed — no post-hoc edits to SQL without re-execution
-
When any shared parameter changes (period, filter, threshold), identify ALL queries using that parameter and re-execute all of them
-
After transcribing results to report tables, verify: bucket sums match totals, percentages match counts/totals, and no stale data from previous executions remains
Interpret Results (Section 4): Analyze query results and draw insights.
-
Interpret results and explain findings
-
Visualize insights with tables and Mermaid diagrams where helpful