Measuring the value
A signed invoice posts itself to the customer's Zalo group: the first PDF arrived within minutes, the ledger corrected before sending
Once an e-invoice is signed, its PDF goes by itself to the Zalo group where the customer already talks, with the amount; before sending, the invoice in the ledger is corrected to match the signed copy. The first one has gone, and only 3 of 20 stores have a group so far. What is this worth, and what assumptions does the figure rest on?
How long does a signed invoice take to reach the customer, and how much of your money does the wait hold?
It needs only the invoices a month, the value of one, the share of customers whose chat group is connected to the assistant, and how many days an invoice takes to reach them today. Everything else is derived. Everything is editable.
Derived (per month)
Assumptions — open to see and edit every number
Money
Time
This is an estimate, not a promise. Set the share of invoices with a chat group to 0: every line goes to 0. Set the days of waiting, the share chased and the share mismatched to 0: it goes to 0 as well.
An e-invoice is signed in the afternoon. The customer is one store of a supermarket chain, and the place where that store talks to its supplier every day is a Zalo group. Until 30/10, the ERP system never knew an invoice had been signed: the job that asks the e-invoice provider for an invoice's status had been written but had never been scheduled to run. A signed invoice sat on the provider's side; by whatever route the PDF reached the customer, the system recorded nothing and counted nothing.
This article asks: if an invoice that has just been signed sends its PDF to the customer's own chat group by itself, with the amount — and, before it is sent, the invoice in the internal ledger is corrected to match the signed invoice — what is that worth each month, and what assumptions does the figure rest on?
1. Which way invoices travel today
Measured on one chain, from 1 to 30/10/2026: 279 invoices for 20 stores, on average about 1.44 million VND an invoice. None of those 279 invoices left any trace in the system of when it was signed or when the customer received it. Across the whole system, as of the morning of 31/10, only 3 invoices had ever been pushed to the provider as drafts, and 1 signed invoice had come back.
In other words, the question «does the customer have the invoice yet» cannot be answered from the data today. The person asking is the receivables accountant, when the due date comes and the money has not; the person answering is the store, by phone.
2. What changes
- Every five minutes, the system asks for the status of the invoices pushed as drafts. Any invoice that has been signed has its XML file and PDF downloaded.
- The signed invoice is compared with the invoice in the ledger, item by item — quantity and amount, a gap of 1 VND counts. Match ⇒ a «matched» note. A gap, and that customer group does not allow the VAT invoice to differ from the internal invoice ⇒ the ledger invoice is cancelled and an amended copy is made with lines exactly as on the signed invoice, keeping the links to the delivery note and the sales order. If it cannot be corrected automatically (an item on the signed invoice does not resolve to a code in the system, or the system refuses) ⇒ the gap is recorded and the invoice is flagged «needs manual adjustment».
- The assistant account sends the PDF to the Zalo group linked to the customer, with one line: invoice number, date, total in VND, delivery note number, the lookup link and the lookup code. Each invoice is sent once (a note on the invoice marks it as sent), to one group — the newest group if the customer has several. A customer with no group connected to the assistant ⇒ nothing is sent, and the reason is noted on the invoice.
- This applies only to the chain whose stores already send delivery paperwork through a Zalo group, and it has its own on/off switch.
The PDF that is sent is the signed copy — a legally valid document. If step 2 cannot correct automatically, the PDF still goes, and the ledger invoice lands on the list needing manual adjustment.
3. What the first hours measured
One PDF sent, on the afternoon of 30/10. That is all the data there is: one out of one signed invoice took this route — one real run, not yet a rate.
But that very invoice shows why step 2 must come before step 3. Before signing, the accountant changed the price of one line on the signing page: that line's unit price fell by about 22%. The signed invoice therefore came to about 13% less than the ledger invoice. Had the PDF gone out with the ledger unchanged, the store would hold one invoice while the receivables ledger claimed 13% more — a reconciliation already booked in advance. The system made an amended copy equal to the signed invoice, and only then sent it. The message reached the group 4 seconds after the system saw the signature; because status is checked every five minutes, the time from signing to the customer having the PDF is under five minutes.
Coverage is the big number today: 3 of the chain's 20 stores have a Zalo group connected to the assistant account. Those three stores account for 36 of the 279 invoices in October — 12.9%. For the other seventeen stores, a signed invoice still goes nowhere by itself.
4. Source one: money no longer waiting on paper
Some customers count the payment term from the day they receive the invoice, not the day it is signed. Each day the invoice has not reached them is one more day the receivable sits on the books.
How to measure. Take one quarter of invoices: signing date, date the customer confirmed receipt, date the money arrived, and the payment term in the contract. A customer who pays exactly at «receipt date + term» is counting from receipt; for a customer who pays exactly at «signing date + term», this source is 0.
Confidence: low to fair. The whole source rests on a question you have to ask your own customers. And money in earlier is one-off: it sits outside the monthly total, do not add it to every month.
5. Source two: nobody has to call asking for the invoice
How to measure. For two weeks, each time someone asks «what's the invoice number, can you resend it», log one line: who asked, which invoice, how long it took. Count both sides if you like, but only cost your own side.
Confidence: moderate. The number of times is easy to count; the minutes each time vary widely by person.
6. Source three: the invoice sent out matches the ledger
How to measure. Count the signed invoices whose total differed from the ledger invoice over three months. The system now notes «matched» or «adjusted» on each signed invoice, so from here on this number produces itself.
Confidence: low, until there are a few dozen invoices. The 3% is an illustration; the single invoice measured did differ, but one invoice is not a rate.
7. Adding it up
| Source | Coverage today (13%) | All 20 stores |
|---|---|---|
| Money no longer waiting on paper | 0.06 m | 0.47 m |
| No calls asking for the invoice | 0.04 m | 0.28 m |
| Invoice sent out matches the ledger | 0.03 m | 0.25 m |
| Per month | 0.13 m | 1.00 m |
| Receivables in earlier, one-off | 6.1 m | 47 m |
At today's coverage, this is worth about 130,000 VND a month — next to nothing. At all 20 stores, about 1 million VND a month, 12 million a year, plus 47 million VND of receivables in earlier, one-off. The gap between the two columns is not in the source code. It is in seventeen chat groups not yet connected. The most worthwhile work this week is connecting groups, not writing more features.
8. What assumptions this ROI rests on
The person who needs the invoice is in the group. A store's Zalo group is where store staff and delivery staff talk. The chain's payables team may sit at head office and never read that group. If so, the PDF reaches the right group but the wrong person, and source one is 0. Ask once: who on the customer's side holds the invoice to pay it, and where do they receive invoices.
The customer counts the term from receipt. As said in section 4: this is the heaviest assumption, and it is a question, not a calculation.
The ledger invoice is corrected automatically. When step 2 cannot correct automatically, the PDF still goes, and source three is 0 for that invoice — it becomes a task on the manual-adjustment list.
The system has to know the invoice was signed. This whole chain rests on a five-minute schedule that did not exist before 30/10. A background job that silently stops will stop everything behind it, with no error. The number to watch each week: signed invoices with neither a «sent» nor a «no group» note — it must be 0.
9. Measure it in your own company
- Take a baseline. One month: invoices per customer group, signing date, date the customer confirmed receipt, date the money arrived, calls asking for invoices, signed invoices differing from the ledger.
- Connect groups before switching on. Count the customers with a chat group that is actually in use; that is the ceiling on every number above.
- Run for four weeks, recording each week: invoices signed, invoices sent to a group, invoices adjusted before sending, invoices needing manual adjustment, calls asking for invoices.
- Compare customers with a group against customers without one in the same month, same chain — that is the fairest comparison you have.
The assistant account that sends the PDF here is also the account working in the chat group in the article an accounting assistant right in the group chat.
10. What to remember
An invoice that has not reached the customer is an invoice whose payment has not yet begun. Digital signing has been fast for a long time; the slow part is the stretch between the signature and the person who needs to hold the invoice. Send the signed copy to the place the customer already talks, and before that, correct your own ledger to match that very document.
The invoice you send the customer and the invoice in your own ledger must be one and the same. If they differ, the reconciliation already has a date.
The number of invoices, stores, stores with a group, invoices pushed and signed, the 13% gap on the first invoice and the 4-second delay were measured on the live system of a fresh-produce supplier to supermarket chains on the morning of 31/10/2026, anonymised. Days waiting, the share of customers counting the term from receipt, the share needing a call, the share of mismatched invoices, minutes and cost of capital are illustrative figures. Replace them with your own.
Read next
-
Make-up orders from a supermarket chain: one PO in 26, and only the receipt tells
A store raises a PO and receives it the same day to pay back goods already delivered without one. The PO header still says "deliver tomorrow", so a make-up order is easily delivered twice and the goods behind it never invoiced. Measured on 1,258 POs, with a matching threshold chosen from the data.
-
One screen for a delivery crew's day: 43% of notes posted before the truck delivers, and returns nobody can record
The warehouse posts the delivery note as the truck leaves; the driver's screen read it as "done" and locked the returns field. Over 14 days, early-posted notes had a signed slip 1% of the time and a returned quantity 0.4%, against 89% and 15.5%. Reopen the note for the driver, add a vehicle & fuel tab that reads photos with AI — what that saves a month.
-
Exceptions that arrive with a reason: 82 slips to check, but only 68 need someone to think
When invoices to supermarket chains raise themselves, the remaining slips reach the accountant with a reason in words. Read by what has to be done, the reasons fall into three kinds: judgement, wait, one-time fix. Measured on 291 slips, and the price of the minutes spent searching.