Industry related info

After the Microsoft outage in South Africa, ask where your records actually live

Microsoft 365 went down for almost 24 hours last week, and SARS and the City of Cape Town were among those warning users. The credit will not pay for the lost day, so the better question is which parts of your business depend on one vendor's channel.

By Arnaud Brunel — Founder, Brunel Studios18 November 2026 Last updated: 18 November 2026
Industry related info

Microsoft 365 was down for almost 24 hours last week, from 6.20pm on Thursday 12 November to 6.10pm on Friday, according to the preliminary report covered in TechCentral's account of what broke Microsoft. Exchange Online, the email behind Outlook, was hit hardest, and SARS and the City of Cape Town were among the organisations warning users. The SLA credit is the obvious thing to chase. The more useful question for a business owner is what stopped in your own business, and where the record of it was sitting.

The credit will not pay for the day

TechCentral's reading of Microsoft's SLA is clear enough. Below 99.9% availability the credit is 25% of the fees for the affected service, below 99% it is 50%, and below 95% it is 100%. Credits are capped at the month's fees, they are the "sole and exclusive remedy", they exclude lost revenue, and an administrator has to claim them by 31 December. Claim it. But even the maximum, a month of licence fees, does not cover a day of quotes that never went out.

It will also happen again. TechCentral noted that Microsoft 365 also went down in January this year and in October last year. Moving provider does not solve it. Every provider has its bad day.

What stayed with me from the report is how the fault behaved. The change went out in July and ran worldwide "without impact" for more than three months. When it finally broke, the usual rollback tools could not quickly undo code that old. Teams, OneDrive and SharePoint could switch away from the failing gateway and partly recovered sooner. Exchange Online could not switch, "for scale reasons". So email, the channel businesses lean on hardest, had no way around.

Treat the vendor's channel as a doorbell

When we scoped a proximity app for a South African client this year, we met a smaller version of the same problem. For push notifications, Apple's and Google's services are the only delivery routes the platforms allow. You cannot avoid depending on them. So we proposed that the message text never goes in the notification. The notification carries the sender's name and tells the app to fetch the message from the client's own Postgres database. The chat also keeps a polling refresh, so a user who declines notification permission still has a fully working app. Push makes it fast. It is not where the message lives.

The second decision was about quiet failures. We specified a token-based key for Apple instead of a certificate, because certificates expire every year and push would fail silently about a year after launch. The maintenance plan we proposed also put the push send trigger and credential health under monitoring. That is the July change problem in miniature.

This is an app's notification design, not corporate email, and I am not saying it would have survived Microsoft's outage. The principle is what carries over.

What to check before the next one

Make a short list of what in your business depends on one outside service being up: quotes, order confirmations, booking alerts, customer messages. For each one, ask what stops if that service vanishes for 24 hours, and where the record would be. If the answer is "in someone's inbox", that is the weak point. The record should sit in a system you own, a CRM, a booking system or your own database, and email should only be the notification about it. That is a basic business continuity plan for a small business, and it is the same thinking behind designing an app that keeps working when the network does not.

If you are building custom software or paying someone to, ask your developer two things for every feature: what happens when this vendor is down, and how would we know if it had been failing quietly for three months?

Microsoft says it will publish its final report, with a full list of repairs, by 20 November. Nobody will write the list for your own business, and it is much cheaper to write it before the next Microsoft outage in South Africa than during one.

Arnaud Brunel

Founder, Brunel Studios

Arnaud Brunel is the founder of Brunel Studios, a software product studio based in Cape Town. He has spent the last 8 years building digital products for founders and SMEs across South Africa and Africa, working across mobile, web and AI-native platforms.

LinkedIn ↗