Goods received notes and the R1bn Eskom never got

Eskom recorded diesel as received in its books and paid for it. R1 billion worth of that diesel has still not been delivered. Receipt is not something anyone observed. It is a database field, and somebody typed it.

A goods received note is the control that exists to stop exactly this. In most South African businesses of 40 to 200 people it is not really a control, because nothing checks it against anything else.

The amaBhungane investigation by Susan Comrie, carried by Daily Maverick and by News24, puts it in one sentence. Eskom paid for diesel and recorded it as "received", and in some instances that process was "at best a sleight of hand, at worst a fraud".

The detail is worth having, because it is the ordinary version of the failure at an extraordinary scale. Eskom prepaid Lanele Resources R721 million for 40 million litres, booked as received at the Burgan Terminal on 3 and 15 July 2025. Five million litres were actually in the tanks. By the end of June 2026 roughly 40 million litres, now worth about R1 billion, was still outstanding, almost all of it from that one supplier. So the hundreds of millions paid and the R1 billion undelivered are the same diesel, priced a year apart.

Eskom's position is that the fuel was "delivered or available", and that available here means diesel "procured and stored offsite and available on scheduled request".

One caution on the sourcing. Both outlets carry the same investigation, so that is one source presented twice, not two.

Receipt is a claim, not an observation

Every business runs on the same quiet assumption: that when the system says goods arrived, goods arrived. Nobody in the finance function saw the truck. What they saw was a field that said received, and they paid against it.

That is fine when three conditions hold. The field is produced by a person who was standing at the gate. That person is holding a document that matches a purchase order, and the entry cannot be edited afterwards without a trace. It stops being fine the moment any one of those conditions fails.

At Eskom's scale the failure is worth a billion rands. At a 60-person distributor in Isando it is worth a few hundred thousand a year, spread across enough small events that nobody ever adds them up.

Three systems, three versions of the same fact

Walk the chain in almost any operations business and you find receipt recorded three separate times, in three places, by three people who never speak.

• The delivery note. Signed at the gate, often by whoever was nearest, sometimes without counting. It goes into a driver's file or a lever arch box.
• The invoice. Captured into the accounting package by the bookkeeper, days later, from the supplier's document rather than yours.
• The stock movement. Entered into the warehouse system by a storeman, if there is a warehouse system, and if it got done.

Three records of one event. None of them references the others. The accounting package is the one that triggers payment, and it is the one furthest from the gate.

Ask which of the three is authoritative and most businesses cannot answer. That is the whole problem in a sentence.

The three-way match, and why it is usually decorative

The textbook control is a three-way match: purchase order, goods received note and supplier invoice have to agree before payment is released. Nearly every operator knows the phrase. Very few have it running as a machine check.

What happens instead is that a person performs the match by eye, at month end, under time pressure, on documents that use different reference numbers. The purchase order says PO4471. The delivery note says the supplier's own docket number. The invoice says something else again. Matching them is detective work, so it gets done properly on the large ones and waved through on the rest.

Then approval hierarchy gets added on top, because approval feels like control. It is not. A second signature on an unmatched record only means two people believed the field.

What it costs before anyone calls it fraud

Fraud is the headline case and the rarest one. The everyday cost is duller and larger.

Short deliveries that were never claimed, because the claim window closed before anyone reconciled. Stock counts that never agree, so the variance gets journalised every quarter rather than explained once. Suppliers paid twice on a re-sent invoice. Working capital tied up in stock the system thinks exists.

None of that reaches a newspaper. All of it comes out of the same broken link. The record that says something arrived is not connected to the record that says something was ordered, or to the one that says something was paid.

Where the fix actually sits

Not in a new system. In the reference that travels between the systems you already run.

Receipt has to be captured once, at the point it happens, against the purchase order, by someone who can see what was ordered while they are counting what came off the truck. A phone at the gate does this. So does a printed order with the quantities on it, if the entry that follows is keyed against that order number rather than typed fresh.

After that the work is joining the three records so a mismatch surfaces on the day rather than at year end. This is the kind of thing Zorah connects between an existing accounting package and an existing stock system. Replacing either one is a bigger project than the problem deserves, and operations is where the money is actually leaking.

What to do on Monday

Pick three payments your business made to suppliers last month, over R20 000 each. For each one, find the purchase order, the signed delivery note and the invoice, and put the three documents next to each other.

Check that the quantities agree, that the reference numbers connect, and that you can tell who recorded receipt and when. If it takes more than twenty minutes for all three, you have the same exposure Eskom has, at a scale you can still do something about.