Five models, audited end to end.

A sales star schema with a budget sheet on the side: the model most teams start from.

• Date table never marked, so time intelligence silently misbehaves
• Bidirectional filter on Customer
• Budget joined on a text category

50 findings · 7 critical · 6 verified fixesOpen the audit

Budget vs actual from a SAP extract, with FX rates in a spreadsheet and a snowflaked chart of accounts.

• FX rate hard-coded at 1.08 in two measures
• Variance % turned into text, then multiplied
• A debug measure and an exact duplicate left visible

63 findings · 10 critical · 9 verified fixesOpen the audit

Monthly headcount snapshots, a skills matrix joined many-to-many, and personal data sitting in a dimension.

• Headcount summed across snapshots instead of taking the last one
• Many-to-many with a bidirectional filter on Skills
• Birth dates and e-mails exposed to report authors

46 findings · 11 critical · 7 verified fixesOpen the audit

MRR, churn and product usage built on one wide subscriptions table and a live Stripe call.

• A live API key inside a Power Query URL
• New MRR anchored on a hard-coded 2026 date
• Churn rate divided without DIVIDE, no marked-date-safe period logic

48 findings · 8 critical · 7 verified fixesOpen the audit

Inventory snapshots, purchase orders and shipments with two calendars and a supplier bridge table.

• Fact-to-fact relationship Shipments → Purchase Orders
• Two bidirectional filters through the Item Supplier bridge
• Stock cover divides by a monthly average that ignores the filter

48 findings · 13 critical · 7 verified fixesOpen the audit