Written once: the economics of straight-through processing
The back office stops producing the close and starts verifying it.
The first paper in this series named the four numbers an insurance system must move; the second showed how the features underwriters ask for map onto them. This paper takes the third seat in the room: operations. Because the largest single running cost in most delegated-authority businesses is not the front office at all — it is the monthly machinery of bordereaux, tax, settlement and cash, and the one to two full-time people (roughly $150,000 to $250,000 a year) it takes to keep that machinery turning by hand.
The root cause is simple to state. In a manually assembled business, the same transaction is written down four or five times: once in the quoting tool, again in the policy documents, again in the bordereau spreadsheet, again in the tax workings, again in the accounts. Every re-entry is paid for twice — once in the labour of typing it, and once in the reconciliation needed to prove the copies still agree. The monthly close is not really a reporting exercise; it is the bill for all that duplication, falling due at once.
Straight-through processing is the removal of every entry after the first. The transaction is written once, at bind — classified to its risk codes, split for premium and tax, tied to its wording version — and everything downstream is derived from that single record rather than re-created beside it. The bordereau accumulates rather than being prepared, and the settlement report is read rather than built. Cash is not ticked off against a statement; it is matched as it arrives, with tolerance for foreign exchange and rounding, leaving only genuine exceptions for a person.
The close, task by task
| The monthly task today | Under straight-through processing | What remains for a person |
|---|---|---|
| Bordereau preparation | Every bound risk writes its own row at bind — coded, split and complete when the cut-off arrives | Review the reconciled bordereau and sign it off |
| Premium, tax and levies | Stamp duty, GST and levies calculated per transaction by the tax engine, by territory, at bind | Review flagged exceptions; approve the return |
| Broker remittance | Statements ingested and auto-matched to a per-broker receivables ledger, with tolerance for FX and rounding | Chase and resolve the unmatched residue only |
| Settlement to capacity | Settlement reports derived from the same record the bordereau came from — they agree by construction | Verify and release the payment |
| Accounts | Summary journals exported to the general ledger | Ordinary month-end review |
| Audit preparation | The audit trail is generated automatically as part of trading — rule, wording version, user, timestamp | Answer questions; nothing to rebuild |
Where the money goes
The economic consequence is that the operations role changes in kind, not just in size. Under the manual model the back office produces the close: it types, copies, checks and argues the numbers into agreement. Under straight-through processing it controls the close: a complete, machine-reconciled position lands on someone’s desk, and their job is to verify it and handle the exceptions. Production work in the order of one to two full-time people compresses into a control function measured in hours per month — and, importantly, the control is stronger, not weaker. Verifying a close that reconciles by construction is a better four-eyes check than assembling one by hand ever was, because human attention is spent entirely on the anomalies rather than the arithmetic.
For the owner, the saving takes one of two honest forms. A firm at steady state removes cost: the headcount is redeployed or not replaced. A firm that is growing avoids cost: the book can double, and the close does not, because a derived close scales with the software rather than the staff. Either way the deeper effect is the same — the cost to run detaches from gross written premium. That flattening is precisely what makes a delegated-authority business scalable, and it is invisible in any feature list.
There is a second saving, easy to miss because it is probabilistic: errors. Every manual touch carries an error rate, and in this business errors are not clerical — a mis-coded or late bordereau row is a compliance event in front of the party that controls the firm’s capacity. Removing the touches removes the exposure. The binder — the revenue ceiling named in the first paper — is safer under a close that cannot disagree with itself.
The close test
Count the number of times a single policy’s data is touched by a person between bind and cash settled. Every touch after the first is cost, and every touch is a chance to be wrong. The straight-through standard is one entry, zero re-entries, and human attention reserved for exceptions. If a system cannot state its number, assume it is high.
The conclusion
Straight-through processing is usually sold as speed at the front of the business. Its larger prize sits at the back: a monthly close that assembles itself, a cash position that matches itself, and an operations role elevated from typing to control. The people either go, or go up — and the cost to run stops climbing with the book. Written once is more than a slogan about data entry — it is what lets a business scale without hiring in step.
Cuttleflow Systems · Perspective 03 · 33°53′S · 151°16′E · Sydney