Nobody sets out to run their company on a spreadsheet. It just happens. Someone builds a tracker for one job, it works, and five years later that file is how the company decides what to buy, what to ship and who to invoice.
The useful question isn’t “are spreadsheets bad.” They’re not. The question is: where exactly does yours stop working, and how will you know you’ve gone past it?
There are four points, and they’re easier to spot than people expect.
One: two people editing at once
This one’s simple arithmetic. One person editing a file is fine. The moment a second person needs to change it on the same day, you have a problem, and there are only bad answers to it.
Cloud spreadsheets push this point back, they don’t remove it. They let people edit at the same time without giving you any way to lock a row, check what was typed, or see who changed a cell and why. The failure is quiet: two people fix the same row differently, the last one wins, and neither of them knows.
The warning sign: somebody in your company says “don’t touch the file, I’m in it.”
Two: around 5,000 rows
The exact number depends on your computer, but the behavior doesn’t. Somewhere in the low thousands of rows with live formulas, three things happen together. The file gets slow enough that people avoid opening it. Filters and sorts get applied and left applied. And nobody reads the whole sheet any more.
That last one is the real problem. Under a few hundred rows, a person can scan the sheet and spot something wrong. Past a few thousand, nobody ever sees the whole thing. Mistakes stop being caught by eye, and nothing else is catching them, because a spreadsheet doesn’t check anything. A cell that should hold a quantity will happily accept the word “backorder” and say nothing.
The warning sign: you have rows nobody has looked at in a year, and you wouldn’t bet money on what’s in them.
Three: the first lookup between files
The moment one sheet reads from another, you’ve built yourself a database with none of the safety features. No way to keep records in step, no way to undo a half-finished change, no backup plan. You also now have a data model — you just haven’t written it down.
This one’s worth watching because it’s where the upkeep stops being proportional. Every new file added to the web multiplies the ways the web can break. Rename a column in one place and a formula somewhere else quietly breaks. Nothing shows an error. A number just goes wrong.
The warning sign: somebody has to open the files in a particular order for the numbers to come out right.
Four: the file has a keeper
This is the one that actually costs money. It arrives when the spreadsheet holds logic only one person understands — a pricing rule buried in a nested formula, a column that has to be cleared every Monday, a tab that looks unused but isn’t.
At that point the spreadsheet isn’t a document any more. It’s an undocumented piece of software with one maintainer, no version history, no tests and no backups, running part of your business. If that person leaves, you can’t hire a replacement, because the job description is “know this file.”
The warning sign: one person’s vacation changes what the company can get done that week.
What crossing one should actually trigger
Not a big system replacement. The most common mistake we see is jumping from “the spreadsheet is struggling” to “we need an ERP”, which is a two-year project to solve a two-week problem.
Cross one or two, and the answer is usually a small purpose-built tool: a real table, a form that checks what you type, a list view and a login. That’s days of work, not months, and it clears up the “who’s editing” and “is this even a number” problems outright, while leaving the process exactly as your team runs it.
Cross three, and the useful move is deciding which system owns which piece of information. You don’t have to build anything yet — just answer “where does the true price live?” and “where does the true stock number live?” Half the tangle of files usually disappears once those have answers.
Cross four, and the priority isn’t efficiency, it’s risk. Get the logic out of that one head and into something written down and testable, even if the screen it lives behind is plain.
Keep the spreadsheet where it’s brilliant
Spreadsheets are still the best tool ever built for exploring a question, trying out a scenario, and doing a one-off piece of analysis. Those uses have no breaking points. Nothing piles up.
The trouble only starts when a spreadsheet stops answering a question and starts holding a record — when it becomes the place the company looks to find out what’s true. Records need rules, history and access control. Spreadsheets have none of the three, by design.
So run the four warning signs. If you can name the person who owns the file, you already have your answer.