Measuring the value
Invoicing what the supermarket chain actually received: 72% automatic, and why the tallest bar waits for a person
Invoicing from the delivery note would be wrong on two invoices out of three, because the chain pays on what it received. When the delivered quantity, the signed handover note and the chain's own receipt agree, the invoice raises itself at the received quantity; the rest reach a person with a reason. What it is worth.
How many delivery slips become invoices on their own, and why do the rest wait for a person?
Enter delivery slips to the chains per month, value per slip, minutes to raise one invoice by hand and the cost of an hour. The automatic share and the reasons come from a real measurement; click a bar to assume that reason is gone.
Derived per month
Assumptions — open to see and change each one
Days waiting for the invoice
Invoices sent back by the buyer
This is an estimate, not a promise. Line C is the softest — set the two sent-back shares equal and read it again. Set the invoicing minutes to 0, the three day fields equal and the two sent-back shares equal: the total goes to zero. An honest model has to be able to.
Every delivery slip to a supermarket chain carries three numbers on each line: what was delivered, from the delivery note; what is on the signed handover note from the store; and what the chain received, from the chain's own goods receipt. The chain pays on the third.
At a fresh-food distributor we counted 209 invoices raised automatically at the received quantity: 136 of them (65%) differ from the delivered quantity on at least one line. 310 of 1,051 lines were invoiced below what was delivered, none above; at the same unit price, what the chain received was 4.7% below the value delivered. In other words, the most obvious automation — invoice from the delivery note — would be wrong on two invoices out of three.
The old way avoided that error by waiting: invoices were raised in the accounting package once the received quantities were known, 3–4 days after delivery (slips delivered on 16–25 October already had 358 invoices there; slips delivered on 26–29 October had 2).
This piece asks: if the invoice raises itself the moment the three numbers agree, at exactly the quantity the chain received, what is it worth — and which slips should still go to a person?
1. What changes
After every run that fetches receipts, the system examines each delivery slip without an invoice. A slip is eligible when:
- every delivered item is on the chain's receipt;
- the price agrees with the purchase order (for a chain whose receipt carries no price, this check is skipped);
- the chain did not receive more than was shipped;
- no line is off by more than a threshold between delivered and received — the threshold is a setting, currently 20%;
- the signed handover note is not in dispute. The note is read by AI from the driver's photo, as described in the signed delivery note, checked the same day.
When eligible, a draft delivery slip is posted and the invoice is raised and posted at the received quantity, with nobody clicking. When not, the slip goes to a review screen with its reasons in words, and the accountant approves, corrects or dismisses it. A slip with no receipt yet waits — for a chain whose stores often enter receipts late, it waits a few days before being treated as "delivered but not received".
Snapshot at 21:40 on 30 October 2026, the first day it was switched on: 209 slips became invoices with nobody clicking (none of those invoices carries an approver's name), 82 waited for a person, 71 waited for a receipt. Another 42 slips delivered before the date invoicing moved into the system were set aside for a person to check and are not counted here. Of the slips with a verdict, 72% became invoices by themselves and 28% waited for a person. A read-only trial week before that (21–28 October) gave 52% on 335 slips. The two figures were measured on different sets; do not read them as a trend line.
2. Why 82 slips wait for a person
A slip can carry several reasons. Counted by slip: 56 have a line off by more than 20%, 18 have goods delivered that are not on the receipt, 12 have an item not yet mapped to a product code, 9 have a price difference, 9 have a handover note in dispute, 6 show the chain receiving more than was shipped.
The Pareto chart in the calculator above draws exactly these numbers, and its line is computed from the real combinations of reasons, not by stacking the bars. Two things are worth reading from it:
- The reasons you can fix once are small. Mapping product codes and correcting the price list release only 6 of the 82 slips. Worth doing, but it does not change the picture.
- The tallest bar has to stay. If the "off by more than 20%" reason went away, 38 slips would invoice themselves. But those are slips where the chain received far less than was delivered — exactly where a person must look: spoiled goods, a wrong delivery, or a wrong number from the chain. Lowering the share that waits for a person by widening the threshold switches off the control itself.
3. First stream: hours of invoicing by hand
How to measure it. Time 20 invoices in a row, short slips and long ones. Read the automatic share straight off the review screen: every slip has a status, and an automatic invoice carries no approver's name.
How far to trust it: fairly. The slips and the automatic share are counted; the 6 minutes must be timed.
4. Second stream: the days waiting for the invoice
A note on counting twice: the days a receipt sits on the chain's portal are part of these 3.5 days. The piece on the robot that fetches receipts counts them too; if you add the two pieces together, take only one.
How to measure it. For every invoice of the last three months, take the invoice date minus the delivery date. After switching on, measure the same subtraction, separating automatic slips from approved ones.
How far to trust it: medium. An invoice raised sooner does not mean the chain pays the same number of days sooner — it pays on its own cycle. If the chain's payment cycle is fixed, set the cost of capital to 0 and keep only the other two streams.
5. Third stream: invoices sent back by the buyer
How to measure it. Count the invoices cancelled or adjusted over the last three months, with the reason for each. Only the ones wrong on quantity count here.
How far to trust it: low — the softest stream. Set the two sent-back shares equal and the calculator still shows about 38 million VND a year.
6. Adding it up
| Stream | Per year |
|---|---|
| Hours of invoicing by hand (500 hours) | 30.0 m |
| Cost of capital of 2.7 days waiting for the invoice | 7.9 m |
| Invoices sent back | 12.5 m |
| Total | 50.4 m |
About 50 million VND a year at the example's numbers. Those 500 hours are money or capacity for the same accountant to look after more customers — never both.
7. What this ROI rests on
The chain's received quantity is right. The chain pays on that number, so an invoice at that number is an invoice that gets paid. But the chain is sometimes wrong too — a mis-weighed crate, a wrong item keyed in. Then the line is off by more than the threshold and the slip goes to a person. That is why the tallest bar on the chart has to stay.
20% is a choice. Lower it and more slips wait for a person; raise it and fewer do, and more gaps slip through. The threshold sits in the settings so that the person responsible can change it without touching code — and it should change with measurement, not with a feeling.
Automatic invoices can still be wrong. On the first day, 41 invoices raised by the robot were cancelled the same day. The cancellation rate of automatic invoices is the first number to watch: when it rises, the eligibility rule has gone loose.
8. Measuring it in your own company
- Take a baseline. For the last three months: days from delivery to invoice, invoices cancelled or adjusted over quantity, minutes to raise one invoice.
- Run read-only for a week. The system examines every slip and records a status and reasons, without raising anything. Read the reasons chart before switching on.
- Switch on and measure weekly. Automatic share, slips waiting for a person by reason, cancellation rate of automatic invoices, days from delivery to invoice. Compare week with week for the same chain — a chain that enters receipts late has more slips "waiting".
9. What to remember
Invoicing fast is not hard; invoicing the number the buyer will actually pay is. When the invoice takes exactly the chain's received quantity and raises itself only when three numbers agree, most invoices go on their own — and the rest reach a person carrying a reason worth looking at.
Automation is not measured by how much the machine does. It is measured by how much it does that nobody has to do again.
The automatic invoices, the slips waiting for a person and the reasons on each, the share of invoices differing from the delivered quantity, the delay of invoices raised in the accounting package and the read-only trial were measured on a fresh-food distributor's live system on 30/10/2026 and anonymised. Slips per month, value per slip, minutes, wage, cost of capital and the share of invoices sent back are illustrative. Replace them with your own.
Read next
-
Never invoice twice: ask the accounting ledger before creating — 199 of 209 invoices were already there
When invoicing moves from the accounting package into the ERP, there is a stretch when both create invoices. If the ERP asks the accounting ledger before it creates one, how many duplicates does a cutover avoid — and what is each duplicate you never clean up worth?
-
Sharing an estimate by one link: what the mailed-Excel loop costs, and "which version is the latest?"
An estimate passes through 2–4 people as Excel files sent by mail, so someone always asks "which version is the latest?". When the whole set — drawings, quantity table, where every number comes from — sits behind one shared project, with people invited by email or sent a view-only link that needs no account, how many hours a month does the review loop get back?
-
Invoicing at the original PO price: the delivery note says 21,000, the chain's PO says 30,000
A retail chain pays the price on its own PO, not the price on the supplier's delivery note. When a machine reads the PO price, invoices at it when the gap is plausible and corrects the price list for next time, what is that worth a month — and which lines still need a person?