The root cause is almost never exotic
A mistyped command, an unreviewed config push, a single untested server — nearly every incident here traces back to an ordinary change that skipped a safeguard, not a sophisticated attack.
Case studies
Real outages. Real, sourced numbers.
Every calculator on this site models a hypothetical incident. These pages walk through seven real ones — what broke, how long it lasted, and the dollar figures companies disclosed afterward, each cited to its public source.
Scroll for the case studies
7
Named incidents, 2012–2024
$7B+
Combined publicly reported cost
Pick a case study
Same formula. These are the real inputs.
Each one links to its primary source — an SEC filing, an incident postmortem, or reporting that cites a company's own disclosed figure — not a rumor or a headline estimate alone.
Incident
Date
Reported cost
Patterns
Seven incidents, three repeating lessons.
The details differ, but the shape of the cost is remarkably consistent across industries and decades.
Duration and cost don't scale together
Knight Capital lost $440M in 45 minutes; Fastly's outage lasted about an hour with no disclosed loss at all. What you do, and to whom, matters more than the clock.
The disclosed number is rarely the full number
Direct revenue loss is usually the smallest, most citable slice. Compensation, fines, litigation, and reputational drag tend to arrive later and rarely get one clean headline figure.
FAQ
Case studies, answered.
Questions that come up when using real incidents to make a budget or reliability case.
Are these figures official, audited totals?
They are the most credible public figures available for each incident — often company disclosures in SEC filings or earnings calls — but methodologies differ between companies and analysts, so treat cross-incident comparisons as directional, not exact.
Why is the combined total an estimate?
The $7B+ figure sums heterogeneous numbers — one company's direct loss, one industry-wide insured-loss estimate, one analyst projection — using different methods. It illustrates scale, not a single audited sum.
Can I use these as inputs to the calculator?
Yes — each case study links to the calculator page that best matches its shape (airline, IT, trading, or general), so you can see how the same formula would model a similar incident at your own scale.
Why isn't [some well-known outage] included?
This list favors incidents with a specific, sourceable cost figure over ones that were widely reported but never quantified publicly. See the by-industry calculators for illustrative modeling of any incident type.
Your turn