01. The missing rows that rewrote economic policy for a decade
In 2010, two Harvard economists published a paper that claimed a critical threshold existed: when government debt exceeded 90% of GDP, growth stalled. Their data was bulletproof. Their conclusion shaped austerity policy across the developed world for years.
In 2013, someone finally checked the spreadsheet.
They found that the researchers had excluded five countries from their analysis—not by accident, but because of a simple copy-paste error. The excluded rows contained data that contradicted the entire thesis. When the missing countries were added back in, the 90% threshold evaporated. The growth slowdown was real, but it wasn't a cliff.
The damage: a decade of policy decisions made on incomplete information, affecting unemployment rates, poverty levels, and government budgets across multiple continents. All because someone didn't verify their data range before publishing.
02. The copy-paste that quietly understated risk by billions
JPMorgan's London division was managing a complex derivative position. To calculate the daily risk (Value at Risk, or VaR), traders used a spreadsheet that ingested market data and calculated the maximum loss they could face in a bad day.
Except it didn't calculate correctly.
Someone had copied a formula from one cell and pasted it elsewhere, but didn't update the cell references. The spreadsheet was averaging the VaR result with its own output instead of using two independent data sources. This underestimated risk. They thought they could lose $20 million in a worst-case day. They could actually lose much more.
When the market moved the wrong way in 2012, JPMorgan's loss ballooned to $6.2 billion. The "London Whale" trade became one of the most infamous derivatives disasters in history. A spreadsheet formula error didn't cause the loss, but it prevented the bank from seeing it coming.
03. The ancient file format that vanished 16,000 COVID cases
In fall 2020, Public Health England was tracking COVID-19 cases in the country. They exported their database into an Excel file to analyze the data and track testing trends.
Excel has a column limit. After 16,384 columns, it stops. When the database was exported, rows beyond that limit were silently dropped. Sixteen thousand cases simply disappeared from the dataset.
The truncated data was published. Testing was allocated based on incomplete information. Cases weren't caught because health officials didn't know they existed. It took weeks to realize why the numbers weren't adding up.
The fix: use a database. Or at minimum, verify that your export hasn't silently lost data. The problem wasn't Excel's fault—it was that someone relied on a tool without understanding its limitations.
Rather find out now than in the room?
Submit your model and get a written report with cell-referenced findings before your numbers go public.
Get Your Model Audited →