Data Tool for Compliance Analysts — Review Sensitive Exports Locally

Compliance and internal-audit analysts work with exports that carry real obligations: access logs, transaction monitoring pulls, HR records, privileged user lists. Sending any of it through a SaaS analytics tool creates a second custody problem. PondPilot keeps the file on your workstation and runs SQL against it in the browser.

Nothing Leaves the Endpoint

DuckDB-WASM executes inside the browser. There’s no upload, no server-side processing, no third-party account holding the data. For analysts working inside SOX, PCI, HIPAA, or SOC 2 control scopes, that architecture simplifies the handling story.

Control Testing with SQL

A lot of control testing is SQL-shaped once the data is exported. Segregation-of-duties checks, terminated-user access reviews, privileged-action review, change-approval testing — all translate cleanly.

-- Terminated users with post-termination system access
SELECT
  h.user_id,
  h.termination_date,
  a.event_ts,
  a.system,
  a.action
FROM 'hr_terminations.csv' h
JOIN 'access_logs.csv' a USING (user_id)
WHERE a.event_ts > h.termination_date
ORDER BY a.event_ts;

Any row returned is a finding. Save the query as the workpaper evidence of the test procedure.

Segregation-of-Duties Checks

-- Users who both created and approved the same vendor
SELECT DISTINCT c.vendor_id, c.created_by AS creator
FROM 'vendor_creates.csv' c
JOIN 'vendor_approvals.csv' a
  ON c.vendor_id = a.vendor_id
 AND c.created_by = a.approved_by;

Sampling for Substantive Testing

Pull a reproducible sample for manual review, with a seed you can document:

SELECT *
FROM 'journal_entries.csv'
WHERE posting_date BETWEEN DATE '2024-01-01' AND DATE '2024-12-31'
ORDER BY hash(entry_id || '2024-audit-seed')
LIMIT 60;

Same seed, same sample — the test is reproducible by the external auditor or by next year’s you.

Large Logs, Modest Laptop

Access-log exports can hit tens of millions of rows. DuckDB’s columnar engine handles that on a standard work laptop without a separate database.

Audit-Friendly Evidence

A saved .sql file plus the input export is a clean workpaper: inputs, procedure, and result are all reviewable. No screenshots of a BI tool, no “I did some filtering in Excel” narrative.

No Account, No Logs

PondPilot requires no signup. There’s no vendor-side record of which files you opened or what you queried. That removes a category of questions from every risk review of your tooling.

Works Offline

Install as a PWA and operate on a workstation without outbound connectivity. Useful when the data classification policy says the file cannot touch the network at all.

Start a Review

Open PondPilot — local SQL for sensitive exports.