Selling when the network is down

Concrete kills signal. A power cut kills the router. Every network in Ghana has its bad hours and its bad corners, and a bundle is personal money that nobody wants an app eating quietly in the background.

So the question is not how to keep a connection. It is what your shop does for the two hours it does not have one, and the answer has to be written down before the morning it happens.

There are exactly two ways this goes wrong. The first is stopping, which costs you the queue. The second is more dangerous and much more common: recording the sale in two places.

The phone hesitates, so somebody scribbles the sale in the book as well, meaning to enter it properly later. Sometimes they do enter it later, on top of the one that was already there. Now the day's takings are wrong in a way that takes an hour to find. Pick one place for the sale and stay there, especially when things are going badly.

Nothing dramatic. That is rather the point.

These two words get used as if they mean the same thing and they do not. The difference is the only place where your records are ever genuinely at risk, so it is worth being precise about it.

Saved on this phone means the sale is written into the phone's own records. It survives the app closing, the battery dying and the phone being restarted. It is a real, final sale, and it counts on the day it happened, even if the shop's records only receive it on Thursday.

Backed up means a copy has left the phone and is now with the shop. From that moment the sale is safe from the phone itself: from a cracked screen, a lost bag, a phone that gets sold.

Between the two, the sale exists in exactly one place. That is not a flaw, it is the trade a shop makes to keep selling with no network. It just means the gap should be short and visible, which is why every record carries its own state and why the count of records still waiting is on the screen rather than buried.

One number, every evening: how many records are still waiting to back up. Zero means everything the shop did today is safe somewhere other than a phone.

If Kwame's phone has been holding sales since Tuesday, that is a Wednesday conversation, not a month-end discovery. It usually means his phone has not had signal at the counter for days, and the fix is as small as standing outside for two minutes.

This is the fear that stops shop owners putting a second person on a phone, and it is a fair one. If you and Kwame both sell all afternoon with no signal, what stops the two phones disagreeing?

Two things, and they are both worth asking any app about. Every record carries its own key, so if the phone tries again later the same sale lands once and never twice. And each phone has its own receipt numbers: MAD-D02-000318 is the Madina branch, the second phone, its 318th sale. Two phones can never print the same number, so nobody has to guess which record is which.

It is a short list, and selling is not on it. A connection is needed to get a copy of the records off the phone, to see another branch's day from where you are, to have the shop's figures computed on records from every phone at once, and to get a new person signed in.

Selling, receipts, returns, recording money out and the daily close for the branch you are standing in do not need one. If any of those stop when the signal stops, that is a decision somebody made, not a law of nature.

Everything already backed up stays with the shop. It comes back on a new phone when you sign in again. The only records genuinely at risk are the ones still waiting on a phone that is lost or destroyed before they leave it.

That is a small, specific, checkable risk, and it is exactly why the waiting count is worth glancing at every evening. Paper had the same risk and no counter: a book that goes missing takes every sale in it, and nobody ever knew how many that was.