There is a spreadsheet in your business that matters more than any of your software. It might be the pricing model, the stock forecast, the job tracker, the commission calculator, the one that works out what everyone gets paid. It has been going for years. It has a tab nobody opens. And it is almost certainly the most load-bearing object in the company.
Nobody planned this. Spreadsheets are how businesses solve problems quickly, and quick solutions have a habit of becoming permanent. Somebody needed an answer, built something that gave it, and the thing that was meant to last a fortnight is now in its seventh year and quoted in board meetings.
The trouble is that a spreadsheet acquires responsibility without acquiring any of the protections that responsibility should come with. If you bought software that did the same job, you would ask about access control, audit trails, backups, and what happens when the supplier disappears. Because it's a spreadsheet, nobody asks any of it.
What you actually own
Look at the critical workbook and ask four questions.
Who can change it? In most businesses, anyone with the link. There is no distinction between the person who needs to read a number and the person who is allowed to alter how that number is calculated.
What happens if a formula is wrong? Usually nothing happens, which is the problem. It returns a plausible number and everyone carries on.
Who else could run it if that person were away for a month? Frequently nobody, and the honest version of that answer is that the business would improvise for a month and hope.
Where does the version you're looking at actually live? Often there are four versions in three places, and the authoritative one is whichever was most recently emailed.
Every one of those would be treated as a serious control failure in any other system. In a spreadsheet they're invisible, because a spreadsheet looks like a document rather than infrastructure.
Ask the person who owns the workbook to explain, out loud, what one specific cell does — a summary figure, not an input. If the answer takes more than a minute, or contains the phrase "I'd have to check", that spreadsheet is running on institutional memory rather than on logic anybody can inspect.
Then ask a second person the same question. If they can't answer at all, you have a single point of failure with a name attached to it.
The four failure modes
1. Silent error
A spreadsheet doesn't crash. It returns a number, confidently, and the number is wrong.
The classic causes are mundane. A formula dragged down 200 rows when the data now runs to 240, so the last forty records are silently excluded. A hard-coded VAT rate from a previous year. A SUM range that was correct until somebody inserted a row above it. A lookup that returns the nearest match rather than an exact one, so an unrecognised product code quietly reports as whatever is alphabetically closest.
Nothing alerts you to any of it. There is no test suite, no validation, no second system to reconcile against. Decisions get made on the output for months, and the error is usually discovered by accident — someone notices a figure that feels wrong, digs, and finds that it has been wrong since March.
This is the failure mode that costs the most and gets discussed the least, because by the time it's found, nobody wants to reopen the decisions that were made on it.
2. Single point of knowledge
One person built it, so one person can safely change it. Everyone else can technically open it, but nobody else knows which cells are inputs, which are calculated, and which are load-bearing constants that look like data.
That person is now structurally unable to leave, be ill, or be promoted without something breaking. It's unfair to them — they're carrying an operational risk that isn't in their job description and probably isn't reflected in their pay — and it's dangerous for you.
It also blocks improvement. Nobody else will touch the workbook, so it can only ever be developed at the rate that one busy person has spare attention. Most critical spreadsheets are therefore frozen at whatever they were when the business last had a quiet fortnight.
3. Silent divergence
Someone takes a copy to try something. The copy gets emailed to a colleague, who saves it locally, adds a column, and shares it back. Two people now work from workbooks that agree on the past and disagree on the present, and there's no mechanism to notice.
This one is peculiarly corrosive because it destroys trust in numbers generally. Once two versions of the same report have shown different figures in a meeting, everyone becomes rationally sceptical of all reporting — and the business shifts back to running on opinion, which is worse than running on a slightly wrong spreadsheet.
4. Growth beyond its design
Spreadsheets are built for a scale, and businesses don't stay at that scale. The workbook that handled 40 customers elegantly becomes unusable at 400: slow to open, impossible to review, and full of manual steps that were trivial at the original size.
The response is usually to make it cleverer rather than to replace it — more tabs, more lookups, a macro somebody wrote once. Each addition raises the cost of ever migrating, which is how businesses end up with something nobody would design but nobody can afford to leave.
Why "we'll replace it eventually" never happens
Because it works. That's the whole trap. The spreadsheet is producing acceptable answers today, so replacing it is never this quarter's most urgent job — right up until the week it isn't producing acceptable answers, which is invariably the week you're busiest.
The second reason is that replacing a spreadsheet looks like a big project. It usually isn't. What's big is the accumulated logic inside it, and when someone finally documents that logic, a striking proportion of it turns out to be obsolete: rules for products you no longer sell, adjustments for a supplier you left, a manual correction that compensates for a bug fixed years ago.
The third reason is that nobody owns the decision. The person who maintains it doesn't have the authority to commission a replacement; the person with the authority doesn't feel the pain.
What to do instead of a rewrite
You don't need to eliminate spreadsheets. You need to stop them being the system of record.
Spreadsheets are genuinely excellent at one thing: letting a person think numerically, quickly, in a flexible space. That's worth keeping. What's not worth keeping is the same object being the authoritative store of facts that the business depends on.
The pattern that works: move the facts underneath it into something with rules — a database that draws its numbers from your actual systems, applies the same definitions every time, and can be queried by more than one person. The spreadsheet then becomes a view rather than a source. A wrong formula in it stops being a wrong number in the business, because the underlying figures can be checked against something.
That's a much smaller job than "replace the spreadsheet", and it removes all four failure modes at once:
- Silent error becomes visible, because two things can be compared.
- Single point of knowledge becomes inspectable, because the rules are written down somewhere other than one person's head.
- Divergence stops, because there is one set of numbers to diverge from.
- Scale stops mattering, because the heavy lifting is no longer happening in a file somebody opens on a laptop.
You don't have to choose between "leave it alone" and "replace it". A middle option: keep the spreadsheet exactly as it is, but have the numbers it depends on delivered into it automatically, from one agreed source, on a schedule.
Nothing changes for the person using it. What changes is that the inputs are no longer typed, no longer stale, and no longer arguable — which removes most of the risk for a fraction of the effort.
A reasonable first step
Pick the workbook you'd least like to lose. Write down, in plain sentences, what it takes in and what it produces. Not the formulas — the intent.
Structure it as four short lists: what data goes in and where each item comes from; what decisions or outputs come out; which rules are applied in between; and which parts require a human judgement rather than a calculation. Two sides of A4 is usually enough.
That document is worth having whatever you do next. If you never build anything, you've at least turned one person's memory into something the business owns. If you do build something, you've already done the part that suppliers normally charge you to discover — and you'll get quotes that can be compared, because everyone is pricing the same thing.
Most usefully, writing it down often reveals that the workbook is doing three separable jobs, only one of which is genuinely critical. That's the one to fix first, and it's usually much smaller than the whole.
The honest counter-argument
There's a version of this article that overstates the case, and it's worth naming. Plenty of spreadsheets are entirely fine. A model somebody uses to think through a decision once a quarter needs no governance at all. A workbook that produces a number nobody acts on isn't a risk — it's just a habit.
The test isn't whether it's a spreadsheet. It's whether the business would notice within a week if it were wrong, and whether anyone would be able to tell. If the answers are yes and no respectively, you're not using a tool. You're carrying a risk you've never priced.