K12, R5 and D2

Reporting and Analytics

The number you file is the number the money made. Reporting is where most institutions keep a second set of arithmetic: a return assembled next to the book rather than out of it. Here the trial balance, the safeguarding return and the warehouse all replay the same entries, so there is nothing to reconcile between them.

Returns per regimeGL ties to trial balanceWarehouse from CDC

The returns, by regime.

Named formats rather than a promise of configurability. A regime that is not here is a conversation, not a silent gap.

ReturnWhat it is
FSA071The UK client money and assets return, computed from the safeguarding engine's own daily figures.
REP017The UK payment services return.
CASS 15The monthly safeguarding return, with breaks typed so timing differences are separable from substantive ones.The separation is the requirement, not a nicety, which is why break type is a column in the engine rather than a note in a comment field.
CESOPThe EU cross-border payment reporting obligation, from the same transaction record.
SAR and STRSuspicious activity and transaction reports, raised from a case rather than retyped into a portal.

The general ledger, and why it ties.

The chart of accounts maps onto the ledger rather than beside it, so a journal-level export is a projection of entries that already exist. Period close, accruals and revenue recognition run on the same rows.

An export that does not reconcile to the book it came from is a spreadsheet, and your auditor will treat it as one. This one reconciles because it is the same arithmetic, not a second implementation of it.

LiveDaily safeguarding reconciliationComputing
Safeguarded balances against the requirementComputing
Safeguarding requirement
Held in safeguarded accounts

The run reports a heartbeat. A reconciliation that stops is raised as its own alert by ops.job_heartbeats, because a check that silently stopped running looks exactly like a check that keeps passing.

And the warehouse.

Analytics reads a copy, because a question nobody anticipated should not be able to slow down a payment.

PieceWhat it does
CDC ingestionChange data capture out of the operational database, so the book is never queried by a dashboard.
Dimensional modelsBuilt with dbt, versioned like the rest of the system.
SCD2 dimensionsSlowly changing dimensions, so a report run today for last quarter uses last quarter's customer attributes rather than today's.
Semantic layerOne definition of a measure, so two teams asking for revenue get the same number.

What it guarantees.

Each of these is enforced rather than intended, and each is something you can sit down and try to break on a call. We would rather you did.

  • Every figure in a return is derived from entries, so the return and the book cannot disagree. There is no second place where a number is maintained.
  • The GL export reconciles to the trial balance, and the trial balance reconciles to a replay of the journal.
  • A report run for a past period uses that period’s attributes, not today’s, because the dimensions are slowly changing rather than overwritten.
  • Analytics never touches the operational database. A badly written query cannot slow a payment down.

Bring last quarter's return.

We will compute it from a ledger seeded with your transaction types and show you, line by line, where the two differ.