During historic market volatility, Robinhood's DNS fell over and 12.5 million accounts couldn't trade.

Early March 2020 was, in Robinhood's words, a time of "highly volatile and historic market conditions". On 2 and 3 March, the trading app and website shut down. Every customer was locked out: according to the US regulator FINRA, that was 12.5 million accounts. They couldn't place, change or cancel orders. Email and in-app support were disrupted for parts of the outage too, and there was no live phone line.

Robinhood's founders explained the cause on 3 March. Its infrastructure "struggled with unprecedented load", from record trading volume and record account sign-ups. That led to a "thundering herd" effect, where a crowd of clients all hit a system at the same moment, and that triggered a failure of its DNS system: the service that tells every part of the system where to find the others.

FINRA's later findings said the same in its own words: "a key firm system was overloaded, which caused a cascading failure of other systems." And it added why: Robinhood's parent company had monitored capacity, but "did not adequately consider the firm's significant growth or the potential impact of extreme market conditions".

The cost was real. FINRA found that during the March 2020 outages, some customers couldn't buy stocks that later rose, couldn't sell ones that fell, had stop-loss orders that didn't fire, or watched options expire worthless. Some lost tens of thousands of dollars. Robinhood said it would add redundancy and reduce the interdependencies in its infrastructure. In June 2021, FINRA fined Robinhood $57 million and ordered about $12.6 million in restitution, over the outages and other failings.

Why it matters: Plan capacity for your worst day, not your average one. And watch the quiet shared services like DNS: when everything depends on one of them, its failure is everyone's failure.