Somewhere in every organisation there is a number that nobody can source.
It is in the budget. It is in the deck. Someone quotes it in a meeting and heads nod. Ask where it came from and you get a chain of shrugs that ends, four people later, with a spreadsheet built by someone who left two years ago.
I have watched this happen in real estate, in e commerce, in marketing, and now in software. It is the most consistent pattern I have seen across completely unrelated industries, and it is almost never treated as a serious risk. We audit money. We audit code. We almost never audit numbers.
How a wrong number survives
A wrong number does not survive because people are careless. It survives because of how numbers travel.
An estimate is created under uncertainty. The person who makes it knows exactly how soft it is. They would tell you, if you asked, that it is a rough figure with a wide range and several assumptions baked in.
Then it gets written down. Writing strips the uncertainty. A range becomes a midpoint. A midpoint becomes a figure. The figure goes into a slide, and slides have no room for caveats.
Then it gets quoted by someone who was not there. This is the moment the number changes species. It stops being an estimate and becomes a fact, because the person repeating it has no idea it was ever soft. They did not remove the uncertainty. They simply never received it.
Then it gets defended. Once a number has appeared in a board pack, revising it costs someone credibility. So it hardens. I have seen numbers defended for years by people who were not in the room when they were invented.
Four steps. No villain anywhere in the chain.
The compounding part
A wrong number in isolation is a small problem. Numbers are rarely in isolation.
In my e commerce years, we planned a season on an assumed return rate. The figure was a couple of points off. That sounds harmless. It was not, because the return rate was an input into four other things: how much stock we bought, how we priced, how much warehouse space we committed to, and how we staffed the returns desk.
Each of those decisions used the wrong figure as a certainty. Each one was individually reasonable. Together they produced a season where we were simultaneously overstocked, underpriced, paying for space we did not need, and understaffed exactly where the pressure landed.
Nobody made a bad decision. The system made a bad decision, because a soft number entered it as a hard one and then multiplied.
This is the thing people miss about estimate error. It does not add. It compounds through everything downstream that depends on it. A figure that is ten percent off does not create a ten percent problem. It creates a problem the size of every decision that trusted it.
Where I see it now
In physical assets the same pattern shows up with a longer fuse and a bigger bill.
A property is assessed. A condition is estimated. That estimate feeds a maintenance budget, a valuation, sometimes an insurance position and a sale price. If the original assessment was optimistic, every one of those is optimistic, and the correction does not arrive for years. When it arrives it does not arrive as a spreadsheet revision. It arrives as a repair invoice, a failed sale, or a dispute.
The gap between when a number is wrong and when anyone finds out is the real danger. In advertising I could be wrong on a Monday and know by Friday. In buildings you can be wrong for a decade.
That gap is exactly why documentation matters more in slow industries than fast ones. When feedback is quick, bad numbers get corrected by reality. When feedback is slow, the only thing standing between a bad number and a bad decade is whether anyone wrote down where the number came from.
Four habits that actually help
I do not have a system for this. I have four habits, and they are cheap.
Carry the range, not the midpoint. When someone gives me an estimate, I write down the range and the date. A number without a range is a claim. A number with a range is information. It costs nothing to keep both and it changes how people use the figure downstream.
Record the source at the moment of creation. Not the department. The person, the method, and the date. Most bad numbers are not wrong, they are stale. They were correct about a business that no longer exists. A date on a figure tells you when to stop trusting it.
Ask what breaks if this is wrong by half. Not by ten percent. By half. If the answer is nothing much, stop refining it and move on, because you are polishing something that does not matter. If the answer is that four other decisions collapse, you have found a number that deserves real work. Most teams spend their precision on the wrong figures entirely.
Name the decisions that depend on it. This is the one nobody does, and it is the one that stops compounding. When a number is created, list what will consume it. That list is your blast radius. When the number later changes, and it will, you know exactly what to revisit instead of hoping someone remembers.
The honest conclusion
You cannot eliminate wrong numbers. Estimating under uncertainty is most of what management is. The goal is not accuracy, which is often impossible, but traceability, which is always possible.
A wrong number you can trace is an inconvenience. You find it, you revise it, you rerun the decisions it touched.
A wrong number you cannot trace is a permanent resident. It will outlive the person who made it, the project it was made for, and quite possibly the strategy it justified.
Go find one this week. Take a figure everyone in your organisation repeats without hesitating and ask three people where it came from.
If nobody knows, you have not found a number. You have found a belief.
I am Issam Fathi, a technology strategist and the product manager of AssetEye by Dronetjek, based in Tetouan, Morocco. I help companies build, adapt, and grow through technology.
Top comments (0)