Spreadsheets are underrated. They are free, everyone can use one, and they will run a real business further than most software people admit. Plenty of companies should not replace theirs.
But there is a point where the spreadsheet stops being infrastructure and becomes a liability, and it is worth recognizing before it introduces itself.
1. One person is the system
Somebody understands how the workbooks connect, and nobody else does. When they are on vacation, certain things simply cannot be done. When they leave, part of your operation leaves with them.
This is the most serious sign because it is a business continuity problem wearing a productivity costume.
2. You reconcile the same data twice
The same numbers live in the accounting system and in a spreadsheet, and someone keeps them in agreement by hand. Every hour spent on that is an hour spent producing nothing, and the reconciliation will eventually be skipped under time pressure, which is when the two versions quietly diverge.
3. Nobody trusts the numbers without checking
When someone pulls a figure and the reflex is to verify it before acting, the reporting layer has stopped functioning. You are still doing the work of gathering data and no longer getting the benefit of trusting it.
4. Errors reach customers
The wrong rate on an invoice. A scheduled crew that never got the assignment. Any mistake that a spreadsheet permitted and a system with validation would have caught. One of these is a bad day. A pattern of them is a structural problem, and customers form opinions about it quickly.
5. You cannot answer questions about the past
Someone asks why a job was priced the way it was in March. The cell says one number now and nobody knows what it said then, or who changed it, or why. Spreadsheets record state, not history — and once you need history, no amount of discipline in the workbook gives it to you.
6. Growth makes things slower, not faster
Doubling volume should not double administrative hours. When it does, the process is linear in the worst way: every new customer costs a fixed amount of manual work forever. That is a ceiling, and it arrives before you expect it.
7. You have started building software in the spreadsheet
Nested macros. A workbook that takes a minute to open. Formulas long enough to need scrolling. Someone maintaining a script that emails the file on a schedule.
At this point you already have custom software. It is just written in the worst possible language, with no version control, no tests, and no error handling.
What to actually do about it
Recognizing the problem does not mean the answer is a big custom build. In rough order of what we recommend:
Buy something, if your process is standard. If an off-the-shelf product covers most of what you do, use it and adapt. This is the right answer more often than studios like us admit.
Replace one workbook, not all of them. The pain is rarely evenly distributed. Usually one process causes most of the damage. Fix that one, in isolation, and see what changes. This is far cheaper than a platform project and it teaches you whether the diagnosis was right.
Build the layer, not the replacement. Frequently the tools are fine and the problem is that they do not talk to each other, so humans are the integration. What you need is the connective layer, not new systems — a much smaller build.
That last pattern is what Flat OS came out of: a staffing operation running on interlocking spreadsheets that no vendor product mapped onto, because the process had evolved for years around how that specific business works. The answer was not to replace the tools underneath but to build the hub that made them behave like one system.
The honest test
Estimate the hours per week your team spends on manual data movement, reconciliation, and fixing spreadsheet-caused errors. Multiply by loaded hourly cost. Annualize it.
If that number is small, keep your spreadsheets and spend the money elsewhere. If it is large — and add the cost of the errors that reach customers, which is usually the bigger figure — you have a real business case, and now you can size a solution against it instead of guessing.
If you want help figuring out which of those three paths fits, tell us what your process looks like.