Before You Launch, Imagine Everything That Could Break

Instagram: A Case Study

Launches go wrong. That's not pessimism, it's pattern recognition. Every time I work with a founder on a release, we spend time asking one question: what could break?

Not because we expect disaster. Because thinking through the risks ahead of time is the difference between a problem you can handle and a crisis that tanks your credibility.

Instagram's founders learned this the hard way, twice.

When they uploaded their app to the App Store, they did it at midnight. They had one server. Within hours, so many people were downloading and using the app that the server overloaded and the whole thing crashed.

They got lucky. Mobile networks were unreliable enough in 2010 that users blamed their carriers instead of the app. Today, that same failure would generate a wave of one-star reviews and social media complaints before you could fix anything. The app would be dead on arrival.

The founders missed a few things that a pre-launch risk conversation would have surfaced:

The App Store is global. Midnight in San Francisco is mid-morning in Asia. You're not launching to your timezone - you're launching to the world.

If it's on the App Store, people assume it works. Users don't think "maybe their server can't handle the load." They think "this app is broken" and move on.

One server is not a launch plan. It's a prototype setup that you forgot to upgrade.

The second near-disaster came after Facebook acquired them. A Terms of Service update included language that users interpreted as "Instagram can use your photos in ads without permission." People panicked. Some deleted the app entirely. The founders had to publicly apologize and rewrite the terms.

This wasn't a legal problem; the lawyers did their job. It was a communication problem. Nobody asked "how might users read this?" before it went live.

Here's what I do with clients before any launch:

Walk through the user's experience from the outside. Not what you intend, what they'll actually see, read, and assume. Where could they get confused? Where could they get angry?

Ask "what if this works better than we expect?" A flood of users is a good problem, but only if you've planned for it. What breaks at 10x your projected load? What breaks at 100x?

Identify who else needs to weigh in. Legal, customer support, ops - whoever will have to deal with the fallout if something goes wrong. Get them in the room before launch, not after.

The goal isn't to prevent every problem. It's to stop being surprised by the predictable ones.

Previous
Previous

Luck Is Not Market Research

Next
Next

The Lawyer You Can't Afford to Skip