Thursday afternoon, Microsoft’s email went down. Outlook, Teams, SharePoint and a long list of other services stopped working for people around the world, and it didn’t fully recover until the next day. On Friday, the FAA stopped every flight departure in the United States for over four hours. That same Friday, the National Weather Service reported nationwide internet and communications problems, and a few hours earlier a tanker had crashed into a cruise ship docked in New Orleans. It was also Friday the 13th, a day many people consider unlucky.
You could feel people reaching for the connection. Reporters asked officials whether the Microsoft outage and the ground stop were related. Articles appeared lining up all of it under headlines like “What is going on?” Social media did what social media does. One list making the rounds on Threads collected more than a dozen “signs”, including outages, fires and a building collapse. There was even a report of a pterodactyl sighting in Mexico.
Notice when those questions were being asked. On Friday afternoon, the FAA was still saying its technicians were investigating. Nobody knew yet what had caused the ground stop, let alone whether it had anything to do with a Microsoft outage the day before or a tanker on the Mississippi. Even now, as I write this, the weather service is disputing reports of how widespread its own outage was. Nobody had an answer, and plenty of people were already sure.
I’m not immune to this either. That same evening, Amtrak’s website and app started throwing errors, and I immediately assumed it was the ground stop: thousands of stranded air travellers all trying to book trains at once. It’s a reasonable guess, and it may well be right. Amtrak’s own explanations, however, pointed in different directions. One mentioned increased demand, while another blamed a system-wide update. I had a cause in mind before I’d seen any evidence at all.
Why are we so quick to connect them?
Part of it is something called the clustering illusion.
“The clustering illusion is the tendency to erroneously consider the inevitable ‘streaks’ or ‘clusters’ arising in small samples from random distributions to be non-random.” Wikipedia, “Clustering illusion”
The key word there is inevitable. If unrelated bad things happen at random, some of them are going to bunch up. A week with four headline failures isn’t evidence of a hidden cause; it’s what randomness looks like when you happen to be watching. Spread out evenly, one per week, would actually be the suspicious pattern.
Another part is the Interpreter, the part of the brain whose job is to weave whatever is happening into a coherent story. It doesn’t wait for good information before it speaks. Four events in two days is exactly the kind of raw material it loves, and “someone is attacking our infrastructure” is a much more satisfying story than “four unrelated things went wrong”. A single cause feels like an explanation. Coincidence feels like no explanation at all.
None of this is unique to the news. I see the same thing with software teams in the middle of an incident. Two alerts fire within minutes of each other and the whole team piles onto the theory that one caused the other. Sometimes it did. Often it didn’t, and the time spent chasing the link is time not spent on the actual problem.
The twist is that sometimes events really are connected, and that doesn’t get us off the hook.
In July 2024, a faulty CrowdStrike update crashed Windows machines around the world and took down airlines, banks and hospitals in a single morning. Thousands of failures with a single cause. The evening before, Microsoft had a completely separate outage in its Azure Central US region, caused by a configuration change. The two were merged almost immediately, and much of the early blame for the crashes landed on Microsoft. Every machine that went down was running Windows, after all.
Same week, same instinct. One connection was real and the other wasn’t, and the urge to connect felt exactly the same both times. That’s the point. The feeling of certainty tells us nothing about whether we’re right.
So the next time a cluster of bad news lands at once, notice the story forming. Ask yourself what you actually know, beyond the timing. Usually the honest answer is “not much yet”, and that’s a fine place to stay for a while.
The cluster is real. The connection is a story.