Answer Business Questions with Validated SQL and Charts
Turn a business question into a pinned metric definition, checked SQL, the right chart, and a plain finding with confidence and caveats.
Before & after
Same prompt, same agent — the only difference is the skill.
That's a concerning trend. To diagnose it, I'd need to look at your data—specifically, churn rate, new MRR, and maybe cohort breakdown by month. Do you have access to a data warehouse like Snowflake or BigQuery? Or can you export CSV of subscription events (sign-ups, cancellations, downgrades)? If you share that, I can write queries and help you spot the pattern.
MRR DROP ANALYSIS FINDINGS MRR June: 480k | MRR July: 441k | Delta: -39k (-8.1%) ROOT CAUSE BREAKDOWN Churn: +2.3% (historical: 1.8%) → -34k MRR loss Downgrades: Pro→Basic tier: 47 accounts (up from 12) → -5.2k loss New MRR: -2.1% vs May (200 new signups vs 204) → minimal impact CHURN COHORT ANALYSIS Cohort analysis by signup month shows high churn concentrated in 2-month-old cohorts (July 2023): • Retention at 60-day mark: 67% (vs 81% historical) • Reason flagged in support tickets: "Scaling costs, not seeing ROI" DOWNGRADE TRIGGER Pricing change (June 15) moved Pro from 99/mo to 129/mo (30% increase) Downgrades spiked June 17-25 (8 days after change) • Users in tech/startup segment most sensitive • Enterprise segment: 0 downgrades TOP 3 ACTIONS (priority) 1. Win-back: Email to 47 downgraders with "Pro-lite" option at 99/mo (recover 60%) 2. Pause price increase for 30-day cohort (June signups < 60 days old) 3. Analyze support churn tickets → update onboarding PREDICTION If no action: MRR stays -8% (new baseline) With actions: MRR recovery to -2% by August Confidence: High (cohort data + support signals align)
About this skill
"Why did signups drop?" "Does this revenue number look right?" Most AI-written SQL runs fine and still returns the wrong answer. It double counts through a join, compares a partial week to a full one, or blames a feature when the real cause was a channel mix shift. This skill makes Claude work like a careful analyst. It defines the metric first, checks every number a second way, and separates what the data shows from what it only suggests.
What it does
- Pins the metric before writing any SQL: entity, grain, numerator and denominator, time zone, and exclusions such as test accounts and refunds
- Writes CTE-structured SQL that is correct for your dialect (Snowflake, BigQuery, Postgres, Databricks, Redshift), including the date-truncation and NULL traps that differ between them
- Catches join fan-out, grain mismatches, LEFT JOINs that silently become inner joins, and immature or partial periods
- Answers "why did X change" in a fixed order: data-health check, arithmetic decomposition, segment cuts, mix-shift test, and comparison to normal variation
- Recommends the right chart, titled with the finding, with partial periods and known events marked
- Works on a live warehouse connection (read-only by default) or on pasted results and CSV files
What you get
A direct answer with a confidence level, the metric definition, ready-to-run SQL, the validation checks that matter, a chart spec, and the caveats and next query. Quick fixes get quick answers.
Best for
Data analysts, analytics engineers, PMs, and founders who query a warehouse or spreadsheet and need numbers they can defend in a meeting.
SKILL.md preview
---
name: answer-business-questions-with-validated-sql
description: Turns a business question into a pinned metric definition, dialect-correct CTE-based SQL, grain and fan-out checks, a change decomposition when something moved, the right chart, and a stated finding with confidence and caveats. Works against a connected warehouse or pasted results and CSV files. Use when the user says "why did signups drop last week", "write a query for monthly churn by plan", "chart this CSV", "does this number look right", "build an activation dashboard", "my revenue doesn't match finance", "my join keeps double counting".
---
# Answer Business Questions with Validated SQL and Charts
… (sign up to view the full skill)More development skills
View all Development skills →Pandas Data Wrangling Without Silent Errors
Correct, fast pandas code for merging, cleaning, groupby, reshaping, and time series, with checks that catch silent row loss.
Apache Spark Job Tuning and Debugging
Finds why a Spark job is slow or failing from the plan and UI metrics, then gives the code and config fix plus how to verify it.
Reliable Data Pipelines and Analytics Models
Build and repair ETL/ELT pipelines with replay-safe writes, explicit data contracts, and actionable quality checks.
Statistical Tests, Power & A/B Readouts
Get hypothesis tests, effect sizes, confidence intervals, and sample-size calculations with experiment validity checks.