Excel is the most quietly powerful piece of software in most offices. It runs budgets, trading desks, and scientific research, usually without anyone checking the formulas too closely.
That is exactly the problem. Below are eight cases where a single dragged cell, a missing minus sign, or a stray autocorrect changed the outcome of something much bigger than a spreadsheet.
1. Excel kept turning gene names into dates, so scientists renamed the genes
Excel has a feature that auto-converts anything that looks date-like, such as “1-Mar” or “Sept 1”, into an actual date. Convenient for expense reports. Less convenient for genetics.
Human genes have short alphanumeric names, and a number of them look exactly like dates to Excel: MARCH1, SEPT1, DEC1. For years, when researchers uploaded gene lists into spreadsheets, Excel quietly rewrote them into calendar dates. The mistake was invisible unless someone thought to check, and often nobody did before the data went into a published paper.
Eventually the problem became widespread enough that the official gene-naming authority renamed the genes instead of fighting Excel. MARCH1 is now MARCHF1. SEPT1 is now SEPTIN1.
Source: The Verge
2. MI5 collected data on 134 wrong phone numbers after a spreadsheet error
The UK’s domestic intelligence agency once requested subscriber information for telephone numbers connected to its investigations. A formatting error in a spreadsheet changed the final three digits of a set of numbers to “000”.
As a result, MI5 wrongly collected subscriber data relating to 134 telephone numbers that had no connection to its investigations. The material was later destroyed and the formatting fault was fixed.
Source: The Guardian
3. Kodak overstated a severance accrual by $11 million because of extra zeros
In 2005, Kodak overstated a severance accrual by $11 million. The cause was a spreadsheet error: according to a company spokesman, too many zeros had been added to the calculation.
The timing made it worse. Kodak was already reporting heavy losses, and the correction landed as one more difficult headline for a company already under pressure.
Source: MarketWatch
4. A cut-and-paste error cost TransAlta $24 million
In 2003, TransAlta, a Canadian power company, was submitting bids for US power transmission hedging contracts. During the final sorting and ranking of the bids, a cut-and-paste error in an Excel spreadsheet went undetected.
The mistake caused TransAlta to buy more transmission contracts, and at higher prices, than it should have. The cost: about $24 million.
Source: The Register
5. A missing minus sign turned a $1.3B loss into a $1.3B gain
In 1994, a Fidelity accountant left the minus sign off a $1.3 billion loss while performing a tax calculation. A loss became a gain, and the total estimate was off by $2.6 billion.
Fidelity had already told shareholders to expect a specific year-end distribution based on the incorrect number, and had to walk it back publicly once the error was caught.
Source: The Washington Post
6. Barclays learned the hard way that hidden rows aren’t deleted rows
During the 2008 financial crisis, Barclays moved quickly to buy Lehman Brothers’ assets out of bankruptcy. The Excel file documenting the deal, containing 1,000 rows and 24,000 cells, was sent to lawyers just hours before a deadline.
Some contracts Barclays did not want had been marked as hidden in the spreadsheet rather than removed. During the process of converting and reformatting the spreadsheet, that distinction was lost, and 179 additional contracts were mistakenly included in the purchase agreement. After the deal was approved, Barclays had to seek legal relief to remove them.
Source: The Register
7. SUM instead of AVERAGE: the Excel error that understated JPMorgan’s risk
Inside JPMorgan’s 2012 “London Whale” trading debacle was a risk model built in Excel. One calculation subtracted an old rate from a new rate and then divided by their sum instead of their average, as the modeler had intended. JPMorgan’s investigation concluded that the error likely muted measured volatility and lowered the bank’s Value at Risk estimate.
The spreadsheet error did not by itself cause JPMorgan’s $6.2 billion trading loss. The bank’s investigations found much broader failures in trading strategy, oversight and risk controls. But the faulty formula was one reason the model made the portfolio’s risk look smaller than it really was.
Source: JPMorgan Management Task Force report
8. A grad student found an Excel error in one of austerity’s most-cited papers
This one had the widest reach. In 2010, Harvard economists Carmen Reinhart and Kenneth Rogoff published research reporting that countries with government debt above 90% of GDP averaged roughly -0.1% economic growth. The finding became highly influential in public debates over government debt and austerity and was cited by policymakers in the US and Europe.
Three years later, graduate student Thomas Herndon, working with economists Michael Ash and Robert Pollin, tried to replicate the results using the original spreadsheet. They found an Excel formula that did not extend far enough down the column, silently excluding five countries: Australia, Austria, Belgium, Canada and Denmark.
But the spreadsheet range was not the only problem. Herndon, Ash and Pollin also identified selective exclusion of available data and unconventional weighting of country results. After correcting the problems, they calculated average growth for countries above the 90% debt threshold at 2.2%, rather than -0.1%.
Source: Herndon, Ash & Pollin, Political Economy Research Institute
The pattern
None of these are really stories about Excel. They are stories about what happens when a single spreadsheet, unaudited and trusted by default, sits upstream of a decision big enough to matter. A formula, a copy and paste, a dragged range. Nobody meant for it to go this far. It just did, because nobody checked.
If a spreadsheet in your own organization has quietly become a critical system, it may be worth looking at whether it belongs in a proper web application instead. PHPRunner can import that data into a real database and generate a multi-user application around it, complete with permissions and an audit trail.