# Business Data Analysis

> Explain a business dataset with source-backed shape, grain, quality checks, distributions, numeric summaries, descriptive metrics and next questions. Use for read-only analysis; do not rewrite source tables, operate CRM entities or make unsupported causal claims.

- Skill: `prompthon-io/business-data-analysis` (Agent Skill, multi-file: 9 files)
- Install (CLI): `npx skillmds@latest add prompthon-io/business-data-analysis`
- Raw SKILL.md: https://api.skillmd.com/api/skills/prompthon-io/business-data-analysis/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: Prompthon-IO (https://skillmd.com/u/prompthon-io)
- Updated: 2026-09-22
- Page: https://skillmd.com/skills/prompthon-io/business-data-analysis

---


# Business Data Analysis

Own **Analyze** in Lesson 4. Read [safety](references/safety-rules.md), [persistence](references/persistence-contract.md) and the runnable [README](README.md) before executing the selected workflow.

## Workflow

1. Select one local CSV/XLSX/clean.json or a canonical course dataset with rows. Confirm what one row means, date range, population and denominator. If the remote record has metadata only, ask for its explicitly selected local artifact; do not invent rows.
2. Profile shape, column types, nulls, exact duplicates and parse problems. For CSV/XLSX, reuse Structure's reader without applying a rewrite. A schema can clarify dates/currencies, but analysis does not modify input.
3. Summarize categorical counts and numeric min/max/sum/mean/median by unit. Keep contact identifiers out of categorical values in stored reports. Flag invalid, negative and IQR-outlier values for review; a flag is not proof the source is wrong.
4. If stage/value fields exist, calculate stage counts, closed-only win rate and open/won values by currency. State the denominator and row-grain assumptions. No closed deals means an undefined rate, not 0%; never add CAD and USD or invent an exchange rate.
5. Produce Overview, Data Quality, Key Patterns, Business Insights, Recommended Next Questions and Source / Query Evidence. Distinguish descriptive observations from causes or forecasts. If duplicate/grain concerns are unresolved, label totals provisional and return to Structure for review.
6. Save a new analysis report/run with source hash or canonical dataset revision and calculation description. Read back that record. Verify the source hash is unchanged; only report artifacts and analysis records may be new. CRM changes require the separate Operate workflow.

## Shared boundary

Use the course-support client for local or explicitly configured Prompthon storage. Remote organization/workspace, server-derived actor, scopes, revisions and canonical readback are mandatory. The Web App backend remains [a tracked dependency](../course-support/references/backend-dependency.md); never give learners a Neon connection string. See [source notes](references/source-notes.md) for licensing and the original implementation boundary.