A 2024 study published in Frontiers of Computer Science found that 94% of business spreadsheets contain errors. Not minor formatting issues. Actual mistakes that affect the numbers.
If your business decisions rely on spreadsheets, there’s a 94% chance at least one of them is feeding you wrong information right now.
The £6 Billion Copy-Paste Error
The most expensive spreadsheet mistake in history happened at JP Morgan in 2012. An employee building a risk model copied and pasted data between spreadsheets. Several cells contained faulty equations because the paste didn’t work correctly. The bank severely underestimated its risk exposure, leading to approximately £6 billion in losses and fines.
This wasn’t a junior analyst. This was a sophisticated financial institution with extensive controls. A simple copy-paste error contributed to one of the largest trading losses ever recorded.
The pattern repeats across industries:
TransAlta (2003): A Canadian power company lost £24 million when an employee misaligned rows while copying and pasting. The error caused bids to match with the wrong contracts, wiping out 10% of annual profit.
Fidelity Investments (1994): A tax accountant transcribed a capital loss incorrectly, turning £1.3 billion into a gain. The company had to cancel a dividend distribution when they discovered the £2.6 billion discrepancy.
Lazard Investment Bank (2016): While advising on SolarCity’s £2.6 billion sale to Tesla, a spreadsheet error double-counted projected debt. The company’s value was accidentally discounted by £400 million.
UK Public Health (2020): England missed 16,000 COVID-19 test results because an Excel file reached its row limit. The old XLS format caps at 65,536 rows. That limitation delayed critical contact tracing.
Why Your Spreadsheet Probably Has Errors
Research consistently shows the same patterns. According to industry studies, 90% of spreadsheets with more than 150 rows contain at least one significant error. Half of spreadsheet models used in mid-sized and large businesses have material defects that can significantly affect results.
The causes are predictable:
Formula errors: Wrong cell references. Broken formulas after adding rows. Incorrect operations. The JP Morgan disaster came from dividing by a sum instead of an average. A single character difference in the formula.
Copy-paste mistakes: Data misalignment. Overwriting formulas with values. Reference errors when duplicating sheets.
Manual entry errors: Transposed numbers. Decimal point mistakes. Entering data in wrong cells. Kodak’s stock tumbled in 2005 partly because someone added too many zeros to a severance accrual. An £11 million typo.
Version control chaos: Multiple people editing different copies. Changes not tracked. No audit trail of who changed what and when.
Hidden data: Barclays found itself legally committed to purchasing Lehman Brothers contracts they never intended to buy because hidden rows in an Excel file reappeared when converted to PDF.
The Structural Problems AI Won’t Fix
Yes, AI assistants can now help with Excel. Microsoft Copilot and similar tools handle simple tasks well: basic formulas, formatting, straightforward data manipulation. For trivial work, they’re genuinely useful.
But the spreadsheets that cause real damage aren’t trivial. They’re the ones with:
Inherited complexity: A workbook built over years by multiple people, with undocumented logic spread across dozens of sheets. No AI can untangle that without understanding the business context.
Interdependent calculations: Pricing models, financial forecasts, inventory systems where changes cascade unpredictably. Getting one formula right means nothing if you don’t understand how it connects to everything else.
Embedded business rules: The spreadsheet encodes decisions made years ago by people who’ve since left. Why does column J multiply by 1.15 but only for certain product codes? The logic lives nowhere except in that cell.
Performance problems: Workbooks that take minutes to calculate because they’ve grown beyond what Excel handles efficiently. Circular references. Volatile functions recalculating unnecessarily. Array formulas spanning thousands of rows.
Integration requirements: Data flowing in from external systems. Exports feeding other processes. The spreadsheet isn’t standalone; it’s a node in a larger system that needs to work reliably.
These problems require someone who understands both the technical mechanics of Excel and the business logic underneath. AI can help with syntax. It can’t help with architecture.
When Excel Is (and Isn’t) the Right Tool
Excel isn’t inherently bad. It’s powerful for:
- Quick calculations and ad-hoc analysis
- Data exploration before building something permanent
- Simple lists and tracking with appropriate controls
- Personal productivity tasks
Excel becomes problematic when:
- Multiple people need to edit the same data
- The spreadsheet is business-critical and used repeatedly
- It contains complex formulas that others won’t understand
- There’s no documentation of what it does or how
- It needs to scale beyond a few thousand rows
- It’s the only record of important business data
Practical Steps to Reduce Spreadsheet Risk
If you can’t eliminate spreadsheets entirely (and you probably can’t), you can significantly reduce risk:
1. Lock what shouldn’t change. Protect cells containing formulas. Use data validation to restrict input types. Make it hard to accidentally break things.
2. Document everything. Add a “Documentation” tab explaining what the spreadsheet does, what inputs it expects and what the outputs mean. Your future self will thank you.
3. Eliminate copy-paste chains. If you’re regularly copying data between spreadsheets, that’s a process failure. Consider linking sheets, using Power Query or moving to a proper database.
4. Version properly. Cloud-based spreadsheets with version history beat emailing files back and forth. At minimum, use dated file names and keep an archive.
5. Test your formulas. Create test cases with known inputs and expected outputs. Verify the spreadsheet produces correct results. Retest after changes.
6. Know when to graduate. A spreadsheet tracking 50 customers is fine. A spreadsheet tracking 5,000 customers with complex relationships probably needs a CRM. A financial model used for board decisions probably needs proper financial software.
The Better Alternatives
For many business processes, better tools exist:
Customer data: CRM systems provide proper relationship tracking, audit trails and multi-user access without version conflicts.
Financial reporting: Dedicated FP&A platforms offer consolidation, scenario modelling and error checking that spreadsheets can’t match.
Inventory management: Database-backed systems handle stock levels, reorder points and multi-location tracking without the row limit issues that plague spreadsheets.
Project tracking: Project management tools provide collaboration features, status tracking and reporting that spreadsheets require extensive workarounds to approximate.
Recurring processes: If you do the same thing in a spreadsheet every week, it’s a candidate for automation. The time spent building a proper workflow pays for itself quickly.
The Honest Assessment
Most businesses won’t eliminate spreadsheets. But they can ask better questions:
- Is this spreadsheet business-critical? If so, has anyone actually verified it works correctly?
- How much time do we spend maintaining this? Is there a better solution?
- What happens if this breaks? Do we have a backup plan?
- Who else needs access? Is the current setup actually workable for them?
The 94% error rate isn’t fate. It’s the result of using a flexible tool without appropriate controls. With proper practices, or by choosing the right tool for the job, you can be in the 6%.
Sources:
- Poon et al., “Spreadsheet quality assurance: a literature review,” Frontiers of Computer Science (2024)
- PhysOrg, “Study finds 94% of business spreadsheets have critical errors” (August 2024)
- Solving Finance, “The Wall of Shame for the Worst Excel Errors” (December 2023)
- qashqade, “The worst financial services Excel errors of all time”
- Silverfin, “The Hidden Cost of Excel in Accounting” (April 2025)
- FormX, “Understanding Manual Data Entry and Why It’s Time to Shift” (March 2024)
SDB Tech Services fixes the spreadsheets that AI assistants can’t. When you need someone who understands both the technical mechanics and the business logic underneath, get in touch.