The money — a bit over $300,000 — came off a council business-park development, routed through a law firm, paid to her agency, then split to her. A little under half landed with her personally. Every payment looked ordinary. An invoice, an authorisation, a transfer. Nothing in anyone’s system flinched.
That’s the part worth sitting with. The fraud wasn’t buried inside a single transaction. It was the series. And no single transaction could ever have shown it, because each one, on its face, was a normal commission on a normal sale.
Compliance systems are built to check transactions. Is the invoice valid, is the payment authorised, does the paperwork reconcile. They are not built to remember what someone claimed. And this fraud lived entirely in a claim — “I did the agency work on this sale” — false thirteen times across eight years. The falseness only exists across the series. It isn’t visible in any one row of the ledger.
You see the same shape in the old mortgage-ramping cases. An agent tells the buyer the house is worth $500k and tells the bank it’s $600k. Two representations, two audiences, one agent. Pull either statement on its own and it’s just a number in a file. Set them next to each other and it’s obtaining by deception. But nothing in a normal agency or conveyancing stack ever sets them next to each other — the buyer’s file and the bank’s file don’t talk, and no one keeps a record of who said what to whom.
So here’s the thing I keep landing on: the unit of fraud isn’t the transaction. It’s the representation. And representations are exactly what nobody keeps.
The uncomfortable part for an agency
You can be the conduit without being the fraudster.
In the Christchurch case the agency processed thirteen clean-looking commissions and was never charged. Nothing suggested it knew. The law firm that paid out on the false representations wasn’t charged either. Both did their job correctly on every individual transaction.
That’s the whole problem. Correct transaction-level processing is precisely what lets a representation-level fraud pass through. The controls worked. The money moved as instructed. The paperwork matched. And a fraud ran for eight years underneath all of it, until the SFO went looking.
The control that would have caught it isn’t “is this commission calculated correctly.” It’s “did the agency work this commission is claiming actually happen” — and that question needs a memory that spans the whole relationship, not a check that fires once per payment and then forgets.
What I’m actually building toward
Not more checks at the point of transaction. There were plenty in the Christchurch case, and every one passed.
The thing that was missing is smaller and duller than a check. It’s a record of the claim itself. “I did the agency work on this sale” — said once, attached to an invoice, then gone. Nobody wrote it down as a claim that could later be tested. So when the same claim came back twelve more times over eight years, there was nothing to test it against.
That’s the whole gap. Not detection — memory. Every representation an agent makes is a factual assertion someone is relying on: the buyer, the bank, the agency’s own accounts. Right now those assertions get used once and thrown away. If they were kept — just kept, in a form you could line up side by side — most of this would look obvious a lot sooner. The buyer heard $500k. The bank was told $600k. The agency paid out on thirteen sales one person never touched. None of those facts is hidden. They’re just never in the same place at the same time.
The ledger remembers the money. Nobody remembers the words.
Every representation should leave a record that can be checked. Nobody kept the words — so nobody could. — Field Notes, Resaido. Based on SFO v Johnston (District Court, sentenced 20 September 2018); names omitted by editorial choice.