THE PROBLEM
What a hard-coded number is, and why it survives
A hard-coded number is a figure typed inside a formula instead of living in its own cell: =B7*0.062 rather than =B7*$C$4. The spreadsheet calculates exactly the same either way, which is why nobody notices. The difference appears later, when someone asks where 0.062 came from. A rate in its own cell can carry a label and a source beside it; a rate inside a formula can only be explained by whoever typed it, from memory.
These numbers survive because they are invisible in normal use. The cell shows a result, not the formula. The workbook passes from one person to the next, the published rate changes, and the copy buried in the formula does not. Nothing errors. The total is still a number — it is simply the wrong one.
The second habit that produces them is repair under pressure: a formula misbehaves during a close, someone overwrites it with the value it should have produced, and the patch stays for years. That one is worse, because the cell no longer calculates at all.
A hard-coded number is not wrong by definition. It is unverifiable by definition, and that is the problem.
BY HAND
Finding them with what Excel already has
Formula view is the fastest manual pass: Ctrl+` (the backtick key, next to 1) switches every cell from its result to its formula. Digits inside the formulas stand out immediately, and a second press switches back. Scan column by column, because a column normally repeats one pattern — the cell that breaks it is the one to read closely.
Go To Special (F5, then Special) separates cells another way: choose Constants to select every cell that holds a typed value, and Formulas for every cell that calculates. It will not look inside formulas, but it finds the other failure — the formula that was overwritten with a plain value. If a column of calculations shows one cell selected under Constants, that cell was typed over.
Find can help in a narrow way: search within formulas for a specific figure, such as a rate you suspect is stale. What none of these do is read every formula for you and report the literals inside it. For a ten-tab workbook that pass takes an afternoon, which is why it is usually skipped.
THE FASTER WAY
Let a checker read the file
The free spreadsheet checker on this site reads an .xlsx in your browser and lists the numbers written into formulas — along with ranges that stop above the last row of data, formulas that break the pattern of their column, criteria that match nothing, and labelled blocks of constants that no formula reads. The file is never uploaded; the reading happens in the page, and keeps working with the network switched off.
It groups what repeats, so a structural constant that appears in every row of a column is one finding with a count, not two hundred separate alarms. What you get is a short list of places to look, exportable as a CSV, in the order they appear in the file.
WHAT TO DO WITH EACH ONE
Not every literal is a defect
Some numbers belong in formulas. Dividing a month number by 3 to get a quarter, rounding to 2 places, taking half of something — these are structure, not data. They have no source to cite and they will not change. Leave them.
A rate, a threshold, a price or a percentage is different: it came from somewhere, and it will change. Move it to its own cell, label it, note the source and its date beside it, and point the formula at the cell. Do this once per figure, not once per row — one cell, referenced everywhere the figure is used.
A formula overwritten with a value needs its history checked before it is repaired: the value may be hiding a calculation that was wrong. Rebuild the formula, compare it with the typed value, and whichever is right, keep the version that shows its working.