Interactive tool

Treasury System-of-Record Map

The cleanest treasury architectures follow one rule: every data domain has exactly one system of record, and everything else is a consumer. Owner is blank on purpose — the point is to force that one decision per domain. Name the owner, and the tool flags anything that breaks the rule. It's the tool behind the system-of-record guide and the reference architecture.

8 domains with no owner named — that's exactly the gap this map exists to surface.

Data domainSystem of record (one owner)Candidate systemsConsumersIntegration
Markdown table
# Treasury system-of-record map

Rule: each data domain has exactly one system of record. Everything else is a consumer or a copy.

| Domain | System of record | Consumers | Integration |
| --- | --- | --- | --- |
| Cash position & balances | — | Treasury, reporting | Real-time / intraday |
| Bank statements (raw) | — | AR/AP, cash mgmt | Daily import |
| Outgoing payments | — | Banks, treasury | Batch + approval |
| FX & derivatives deals | — | GL, risk, reporting | On deal capture |
| Debt & investments | — | GL, reporting | On deal capture |
| Bank & account master | — | All systems | On change |
| Cash forecast | — | Treasury, FP&A | Daily / weekly |
| Accounting entries (GL) | — | Close, reporting | On posting |

The seeded domains are a common starting point, not a prescription — rename, re-own and extend them for your landscape. The value isn't the table; it's the argument it forces: for each domain, who owns the truth? The export escapes markdown, so a value like ERP | TMS can't break the table (though the tool will flag it as two owners). Nothing you type here leaves your browser.

How this works

Methodology

One row per data domain, one owning system per domain. The tool flags anything that breaks the rule — a domain with no owner, two owners, a flow written in as the owner, or a duplicated domain — and exports a markdown table.

Assumptions

  • Each domain genuinely can have a single system of record; if it cannot, that is a design smell to resolve.
  • The seeded domains are a starting point — you rename and extend them for your landscape.
  • Consumers read the domain; they do not own it.

Limitations

  • Multi-owner and process detection is heuristic — it catches common patterns, not every case; treat flags as prompts, not verdicts.
  • It maps ownership, not the integrations themselves — the integration column is a note, not a spec.
  • It does not verify that your named systems actually behave as the system of record; that is an implementation question.