How do you analyze Paystack payments in Metabase?

{
  "mcpServers": {
    "paystack": {
      "command": "npx",
      "args": ["@paystack/mcp-server", "--api-key", "sk_test_..."]
    }
  }
}

# Install the Metabase CLI
npm install -g @metabase/cli

# Log in (opens your browser; requires Metabase v62+)
mb auth login --url https://your-metabase.example.com

# Load a scoped Paystack export - creates a table and a model
mb upload csv --file paystack-export.csv --collection "Finance & Accounting"

# Refresh the same table later
mb upload replace <table-id> --file paystack-export.csv

Create a polished Metabase dashboard for Paystack finance analytics.
Work end to end: find existing modeled data first; if none exists, load a minimal
read-only snapshot or describe the warehouse sync required.

Goal: Connect Paystack to Metabase for payment success, settlements, fees, refunds, disputes, channels, and reconciliation dashboards.

Inspect before querying:
- Inspect actual tables, column names, currencies, status values, and timestamps.
- Do not assume the source schema matches this guide exactly.
- Reconcile source totals before publishing any finance KPI.

Important:
- Do not claim Metabase connects natively to Paystack; it reads a
  supported database or CLI-uploaded tables.
- Keep native amount, native currency, and reporting-currency amount separate.
- Distinguish authorization, posting, payment, settlement, and reconciliation
  states where the source exposes them.
- Use stable business keys and apply updates/deletes during incremental syncs.
- Exclude account numbers, personal data, and secrets from broad dashboards.
- Only build cards whose required fields and history are actually present.

Dashboard sections:
1. Finance overview: Gross and net payment volume; Payment success rate; Settled vs. unsettled amount; Fees and refunds.
2. Operations: Failed transactions by reason; Pending settlements; Refund queue; Dispute exposure.
3. Controls and reconciliation: Successful charges missing settlement; Settlement amount variance; Duplicate references; Webhook freshness.
4. Planning and trends: Channel mix; Settlement forecast; Fee-rate trend; Currency and country mix.

Suggested models: modeled_paystack_transactions, modeled_paystack_settlements, modeled_paystack_refunds_disputes.

Output: Build the dashboard if you have permission; otherwise provide the exact
questions, SQL, model definitions, and layout. State refresh time, reporting
currency, accounting basis, and any reconciliation caveats.

SELECT
  channel,
  COUNT(*) AS attempts,
  COUNT(*) FILTER (WHERE status = 'success') AS successful,
  ROUND(100.0 * COUNT(*) FILTER (WHERE status = 'success')
    / NULLIF(COUNT(*), 0), 2) AS success_rate_pct
FROM modeled_paystack_transactions
WHERE created_at >= CURRENT_DATE - INTERVAL '30 days'
GROUP BY 1
ORDER BY attempts DESC;

Separate payment success from settled cash and give finance an exceptions queue.

SELECT
  transaction_reference,
  paid_at,
  amount,
  currency,
  CURRENT_TIMESTAMP - paid_at AS unsettled_age
FROM modeled_paystack_transactions
WHERE status = 'success'
  AND settlement_id IS NULL
ORDER BY paid_at;