Mega-class Excel bugs have come to cost companies billions at worst.
Microsoft’s Excel is an excellent tool, but it is also clunky when it comes to formatting. So it’s no wonder that about 90 percent of tables with more than 150 rows contain at least one big error.
These mistakes can have quite serious consequences.
Qashqade says that Tesla was buying SolarCity Corp in 2016. Lazard Investment Bank consulted SolarCity on acquisitions, but there was a small calculation error in one of the Excel spreadsheets it prepared about stock prices. Because of Lazard’s mistake, SolarCity inadvertently gave Tesla a $400 million discount.
In 2014, a similar incident happened to Tibco Software. At the time, Vista Equity Partners was buying Tibco, and financial giant Goldman Sachs was consulting on the sale. However, the Excel table prepared by Goldman Sachs had inadvertently exaggerated the value of Tibco by 100 million dollars. The mistake was noticed in time and the acquisitions went through – but Tibco’s shareholders must have been upset when the one hundred million pot was just an illusion created by the consultants.
Even these are small compared to the worst Excel bug in history, which is in a class of its own.
In 2012, a broker called the “London whale” created a risk model in Excel, where he copied data crosswise from several databases to another. The data that drives great results must have looked good to JPMorgan Chase & Co. until the errors began to surface. Soon the entire risk model was revealed to be one massive mistake that cost the investment giants a total of $2.6 billion in direct losses and $920 million in fines.
Compared to this, the error of the US real estate loan company Fannie Mae is quite small. In 2003, Fannie Mae published its interim report and operating profit – only to discover afterwards that a small error had been made in the invoices, underestimating shareholders’ holdings by $1.1 billion. Even though Fannie Mae tried its best to reduce the damage with an active PR campaign, as a result of the mistake, the value of the company decreased by six percent and Pulju ended up in front of the supervisory authorities to explain the matter.
Even smaller bugs can be quite annoying. Ars Technica tells how in August 2023, the police in Northern Ireland had to apologize for a monumental data leak. The leak in question arose when a citizen asked for information about the police, and the police, according to the law, responded to this. Unfortunately, the table that was openly shared on the internet had another sub-page, and under it, the personal information of 10,000 police officers was found.
In 2021, Wales suffered when a series of Excel errors disrupted the recruitment of anesthetists. The Anaesthetic National Recruitment Office (ANRO), which is responsible for finding employees, had drawn up an Excel-based assessment template that told every single candidate that they could not be hired. Some of the examinees had full marks in the interviews, but the table still gave a red card.
The reason fell into the arms of the centralized data management process. The Excel tables made in different regions were not standardized in any way, so their formatting, ways of handling names and general structure could differ considerably. What’s worse, data was copied from one table to another by hand, which not only took time but also caused a lot of errors. Everything could have been avoided if the table had been standardized and tested, and if at least some degree of verification had been used in processing the results.
That, too, will be lost to Crypto.com in 2021. The cryptocurrency pile accidentally transferred $10.5 million to an Australian customer instead of $100, because a small error had slipped into Excel. The employee who handled the matter inadvertently entered the customer’s account number in the field reserved for the refund amount. The mistake was only discovered seven months later, after which the Australian who received the money was caught trying to flee to Malaysia with the massive pile of cash.
In 2022, Iceland’s state-owned Íslandsbanki put a 22.5 percent slice of its shares up for sale. However, securities were traded at a significant underprice due to an error in the Excel spreadsheet. The error occurred when entries in different tables were piled together, and the data was not cleaned and formatted with due care. There was time to sell a significant amount of shares at a lower price before the brake handle was pushed, and the bank had time to suffer losses of around 19 million euros.