What works, what does not, and since when
We do not promise a service without outages. We promise to say when there is one, without waiting to be asked, and to explain afterwards what happened.
The log is open, and it is empty
The platforms are not yet open to the public. This log therefore begins with no history, and that emptiness is accurate: it does not mean that everything has always worked, it means there has been nothing to record.
From the moment we go live, every interruption that crosses the thresholds set out below is published here while it is happening, then completed with a report. We do not erase a past incident and we do not rewrite it: a correction made to an incident note appears as a dated correction.
The components tracked
Each is monitored separately, because they go down separately. A payment outage does not stop people chatting, and one model going down does not stop people paying.
| Component | What it carries | Status |
|---|---|---|
| Lazao | Conversation, sign-in, history. | Before launch |
| Mobile money collection | Mobile wallet payments, by country. | Before launch |
| Card collection | Bank card payments. | Before launch |
| Inference | Routing to the models. | Before launch |
What triggers a publication
These thresholds determine what we publish and how fast. They count as a commitment: if an incident crosses them, it appears here, whether the cause is ours or a provider's.
| Severity | What it means | Publication | Report |
|---|---|---|---|
| Major | A component is unusable for all users, or a payment has been taken with nothing in return. | 30 minutes | 5 business days |
| Partial | A component is degraded, or unusable in one country or for one payment method. | 2 hours | 10 business days |
| Minor | A perceptible slowdown with no interruption of service. | Mentioned in the monthly summary | — |
What a report contains
Four things, in this order, and nothing more: what users observed, the exact duration between the start and the restoration, the technical cause in plain words, and what we are changing so that it does not happen again.
What a report does not contain: any person's name, and no apology standing in for an explanation. An outage caused by a provider is attributed to that provider, named, because you have the right to know what the service you pay for depends on.
When an incident has led to a charge with no service delivered, the refund is made without any claim from you, and the report says how many accounts were affected.
How you will hear about it
During a major incident, the information appears here first, then in the application, then on our public accounts listed in the footer. We do not send individual messages, except when your account has been affected financially.
The monthly summary on the Transparency page brings together all the month's incidents, minor ones included, with the cumulative downtime per component.
If you see a problem that is not reported here, write to us from the Help page. A report from a user is worth more than monitoring that saw nothing.
No incident recorded to date.