Why a manual counter still beats a queue robot.
The obvious question about topupcove is: why not automate the whole thing? An etisalat quick pay could be a payment gateway that pushes an API call to the operator, all in ten seconds, with no human in the loop.
We tried a smaller version of that. It is fine when everything works. When something goes wrong — a bounced push, a card that mis-flags as fraud, a duplicated request — the customer ends up in a chat with a bot that has no permission to fix anything. The 15-second win becomes a two-day thread. That is the failure mode we want to avoid.
What the 24-hour reply buys
A written record. Both sides have an email trail. The manager confirms the number and the amount before pressing anything. If a fresh cardholder gets flagged by the gateway, we notice inside the reply thread and switch to bank transfer without making the customer redo the whole request.
It also means we do not need to handle chargebacks the same way an automated shop does. A refund is a reply, not a form. It is unglamorous, but the numbers on our refund rate are lower than the ones we saw when the process was “instant”.
What we give up
Speed for the common case. A person who wants 50 AED at 23:00 and expects to see it on the SIM by 23:01 will not love us. We are honest about this: the reply window is 24 hours; the top-up itself is often much faster once you confirm the payment. But 24 hours is the promise.
Why this scales down cleanly
Because the counter is one desk. Doubling the counter would mean doubling the people, not doubling the servers. That is a very small business. It is a deliberate ceiling — it forces us to say no to features that only work at 10× the traffic.
If it stops being enough, we will grow. If it stays enough, we will stay small. Neither outcome would push us to swap the person at the desk for a queue robot.