The spreadsheet that runs your business

It usually appears in the second week of an engagement. We ask how commission is calculated, or how revenue is split across entities, or where the figure on page four of the board pack comes from, and someone opens a workbook with fourteen tabs, links to three other files, and a column of manual adjustments nobody can fully explain. It has been running for four years. One person understands it. That person is on annual leave.

This is not a criticism of the person who built it. That spreadsheet exists because the business needed something the system could not do, and somebody solved the problem. The risk is that a temporary solution became permanent infrastructure without anybody deciding that it should.

Why this is a real risk, not a tidiness complaint

Research into spreadsheet error rates has consistently found that a large majority of substantial workbooks contain at least one material mistake, and our experience testing them is entirely consistent with that. The failure modes are mundane: a formula that stops one row short of the data, an absolute reference that should be relative, a hardcoded value typed over a formula during a late night at year end, a sheet that breaks silently when a source report adds a column.

Layered on top is key-person risk. If one person can produce the number and nobody else can check it, you do not have a control, you have a dependency. Auditors will find it. A lender doing diligence will find it. Occasionally a leaver finds it first.

Find them before deciding anything

• Trace every figure in the board pack back to its source. Any figure that passes through a spreadsheet on the way is on the list.
• Add anything that touches money leaving the business: payment runs, commission, bonus, expense allocations, intercompany recharges.
• Add anything feeding a statutory or tax filing, and anything feeding a covenant calculation.
• For each one, record the owner, the backup owner if there is one, the inputs, and what happens if it is wrong.

Most teams find between five and fifteen. Rank them by the size of the number they touch and the consequence of an error, and work from the top. The list itself is often the most useful artefact of the exercise, because it turns a vague unease into a finite, prioritised set of problems.

Three outcomes, not one

The instinct is to move everything into the system. That is right for some workbooks and wrong for others. In practice each one lands in one of three places. Eliminate it, where the ledger or a connected tool can do the job natively and the spreadsheet is habit rather than necessity. Rebuild it properly, where the logic is genuinely needed but the implementation is fragile: one purpose per file, inputs separated from calculations from outputs, no hardcoded values inside formulas, documented assumptions, version control, and a second person who can run it. Or accept it, where the analysis is genuinely one-off or exploratory, in which case it should never feed a reported number without review.

Spreadsheets are not the enemy

The goal is not a spreadsheet-free finance function. Excel remains the best modelling and analysis tool most teams have, and a business that bans it simply ends up with the same logic hidden somewhere less visible. The goal is that no recurring, material number depends on an undocumented file that one person understands. That is a control question, and it is answerable in a few weeks of deliberate work.