Spatium Flow  ·  Product 01  ·  Flow Audit

The money was never lost. It leaked at the seams between systems.

What was sold, what was contracted, what was billed, what was paid out — four systems that should agree and quietly don't. Flow Audit reconciles them and prices every gap. Here is one run against a 500-account synthetic book.

€4,619,320
Total leakage & revenue-at-risk detected
267
discrete findings
21 dbt models 52 data tests 100% precision & recall DuckDB · BigQuery-portable
01

What leaked

Six leak types, each with a euro figure, a count, and an impact class a non-technical manager can act on. Click any card to inspect the actual flagged records.

Revenue at risk — forecast that may not convert Overstated revenue — books count money that isn't real Recoverable cash — money owed or paid in error
02

How it finds them

Detection isn't pattern-guessing. Each leak type is a precise reconciliation between two sources of truth, netting out anything finance legitimately approved.

01
Land & stage the four systems
CRM opportunities, contracts, invoices, and commission ledgers are staged into typed, tested models — one clean grain per source.
02
Reconcile expected vs. actual
Intermediate models compute what should have happened — expected billing from contract terms, expected commission from plan rate — then compare against what did.
03
Net out approved exceptions
Finance-approved credits, spiffs and accelerators look exactly like leaks. Detection consults the approval registers; leakage is defined as unexplained variance only.
04
Price it & classify it
Every finding carries a euro impact, a reason_code, and an impact class — overstated revenue, revenue at risk, or recoverable cash.
05
Audit the audit
A calibration model joins findings against a ground-truth ledger and computes precision and recall per leak type — so the claim is falsifiable, not asserted.