All that data sitting in spreadsheets and exports? It has answers you haven’t asked yet.

Overview

Take real data from your business or personal life, design a schema, build a database, and create a queryable dashboard. This is Module 07’s three-questions framework put to work.

Draws from: Modules 06, 02 Time estimate: 3-4 hours Difficulty: Medium Prerequisites: At least 2-3 data sources you care about (spreadsheets, CSV exports, accounting data)

Step-by-Step

Hour 1: Gather and Assess Data (30-45 min)

1. Export data from 2-3 sources. Possibilities:
   • Accounting tool (QBO/Xero → CSV export)
   • Time tracking (Toggl, Harvest → CSV)
   • Project management (Asana, Notion → CSV or API)
   • Client list (CRM, spreadsheet, contacts)
   • Revenue/sales data (Stripe dashboard export, Shopify export)
2. Open each file. Assess quality using the Module 07 “Would I Trust This?” test:
   • Are there obvious duplicates?
   • Are formats consistent?
   • Are there blank fields that matter?
3. Make a list of the questions you want to answer with this data:
   • “Which clients generated the most revenue last quarter?”
   • “How many hours did I spend on each project this month?”
   • “What’s my effective hourly rate across all projects?”
   • “Which months had the highest expenses?”

Hour 2: Design Schema and Build Database (45-60 min)

1. Describe your data to Claude using the three-questions framework from Module 07:
   • What are the nouns? (clients, projects, invoices, time entries…)
   • What do you know about each noun?
   • How do the nouns relate to each other?
2. Ask Claude to design a SQLite schema
3. Review the schema — does it capture the relationships correctly?
4. Have Claude write import scripts for your CSV files
5. Run the import and verify row counts match

Hour 3: Validate and Query (45-60 min)

1. Run validation queries:
   • “Are there any duplicate clients?”
   • “Show me entries with suspicious values” (negative amounts, impossible dates)
   • “Which records have missing required fields?”
2. Fix any issues found
3. Start asking your real questions:
   • Start with simple aggregations (“total revenue by month”)
   • Move to more complex analysis (“revenue per hour by client”)
   • Try cross-referencing sources (“projects with time entries but no invoices”)

Hour 4: Visualize (45-60 min)

Choose your path:

Quick path (Claude-generated charts):

• Ask Claude to generate HTML/Python visualizations for your key metrics
• Save as standalone HTML files you can open anytime

Lasting path (Metabase):

• Install Metabase via Docker: docker run -d -p 3000:3000 --name metabase metabase/metabase
• Point it at your SQLite database file
• Build 3-5 dashboard cards for your key questions
• This becomes a persistent, auto-refreshable dashboard

What Success Looks Like

• A SQLite database with clean, validated data from multiple sources
• Answers to at least 5 real business questions you couldn’t easily answer before
• Either standalone visualizations or a persistent Metabase dashboard

Stretch Goals

• Add an MCP server connection so Claude can query your database directly in future sessions
• Build an automated data refresh: export → import → update cycle
• Add a data quality check that runs on each import

Previous: Project A: Upgrade Your Claude Code Setup | Back to Project Lab | Next: Project C: Set Up OpenClaw with One Real Skill