Excel Replacement Assessor

What this tool does

CM8-336 takes the spreadsheets your business actually runs on and scores each of them across four dimensions: how many people share it and how, how fragile its logic is, what happens if it is wrong, and what it costs to keep alive. It then gives a defensible recommendation — leave it alone, fix the controls, replace it with a purpose-built tool, move it to a database-backed application, or accept that the job belongs in a proper system.

It exists because the argument about spreadsheets is almost always conducted badly. One side says "it works fine"; the other says "spreadsheets are dangerous". Both are right about different files, and neither position can be tested without asking a consistent set of questions about each one.

Everything runs inside this single file: no account, no upload, no network request, which matters when the output is a candid list of where your organisation is exposed.

Why spreadsheets are the right answer, until they are not

A spreadsheet is the fastest way in existence to get from a problem to a working answer. It needs no project, no budget, no supplier and nobody's permission, and the person with the problem can build it themselves. That is a genuine engineering achievement and the reason spreadsheets have outlasted every product built to replace them.

The failure is not the tool. It is that a file which was the right answer for one person solving one problem is quietly promoted, over several years, into a system that six people share, that the month-end depends on, and that nobody has understood since its author left. Nothing about that promotion is ever decided. This tool exists to make the decision explicit, once, per file.

The four dimensions

Each contributes up to 25 points.

Sharing and version control is about concurrency. A spreadsheet is a file, and a file has one writer. The moment two people need to change it at once you are running a database with no locking, no transactions and no audit trail — and the failure mode is not an error message, it is two plausible versions of the truth. Emailed copies merged by hand is the single most reliable predictor of a wrong number reaching a decision.

Fragility of the logic is about whether the file can survive being changed. Long nested formulas, array formulas, helper sheets, macros and links to other files are all things that work perfectly until somebody inserts a row. Macros deserve their own weight: code in a spreadsheet is software written without tests, review, source control or documentation.

Consequence if it is wrong is deliberately about impact rather than likelihood. A simple, well-built file that decides something important still scores here, and that is correct — the score is telling you what is at stake, not predicting failure. Personal data, financial reporting, external regulation and the absence of a backup all add to it.

Maintenance and errors is what it already costs. Hours spent fixing, unpicking, merging and explaining; errors that actually escaped; and whether one person is the only one who understands it. Treat a reported zero errors with suspicion — it usually means nobody checks.

The recommendation bands

• Leave it alone. Doing its job, low stakes, cheap to run. Replacing it would cost more than it could ever save, and a large part of a healthy estate should sit here.
• Keep it — fix the controls. The right answer far more often than people expect. One cloud copy with version history, locked formula cells, input validation, a documented independent check and a tested backup will move most files down a band for a few hours of work and no budget.
• Purpose-built tool. The job is well defined but the file can no longer carry it. Usually a small application rather than a platform.
• Database-backed application. Concurrent editing, history and an audit trail are doing real work, and no amount of care applied to a file provides them.
• Belongs in a proper system. At this level the function is almost always one your existing systems already claim to perform. The first question is not which product to buy — it is why the one you already own is not being used, and the answer is usually a configuration decision made years ago by somebody who has left.

All four thresholds are settings. If your organisation's appetite differs, move them — but move them before you assess, not after you dislike an answer.

The control discount

Ticking "has meaningful controls" reduces the total by a fifth. It is the only discount in the tool, because it is the only improvement a team can make without a project.

Meaningful means: formula cells locked so they cannot be typed over, validation on the inputs, and somebody other than the author checking the output against something independent. A password on the file is not a control. The discount is capped at a fifth deliberately — controls on a fragile file are still controls on a fragile file, and no amount of process turns a shared workbook into a system with an audit trail.

Costing the do-nothing option

Maintenance hours are the most useful number on the form and the one nobody has. It is the time spent fixing the file, unpicking what somebody did to it, re-merging copies, explaining it to a new starter and rebuilding it after a column was sorted without the rest — not the time spent using it for its purpose.

Priced at the rate in Settings, it becomes the cost of doing nothing, and it recurs every single year. Set against a replacement quote, it is usually more persuasive than any argument about risk, because it is a number the finance function already recognises. If you do not know the figure, log it for one month on the worst file. That log is the business case.

Before you start a replacement project

A high score is a case for looking properly. It is not an instruction to buy software, and this needs saying plainly because the tool would otherwise be used to justify exactly the projects that go wrong.

Replacement projects fail more often than spreadsheets do. They fail because the spreadsheet encoded years of accumulated exceptions that nobody documented, because the new system does ninety per cent of the job and the missing ten per cent is the reason the spreadsheet existed, and because the people who understood the process were not involved. Before committing: find out what the file actually does — including every exception; check whether a system you already own does it; and price the replacement against the maintenance cost above rather than against a feeling.

And be ready for the answer to be "fix the controls and leave it". That is a good outcome, not a failure of nerve.

The scoring, in full

FAQ

A simple file scored high. Why? Almost always the consequence dimension. Something small and simple that feeds statutory reporting, holds personal data and has no backup is exposed regardless of how neat its formulas are. Look at the dimension breakdown rather than the total.

We have hundreds of spreadsheets. Where do we start? Do not assess them all. Ask each department for the three files they would panic about losing, and assess those. The estate has a long tail that genuinely does not matter, and assessing it is its own kind of waste.

Can the score justify a purchase? It can justify investigating one. The purchase needs a business case of its own, and the maintenance cost from this tool is a legitimate input to it.

Everything scores above 45. Check the criticality answers first — if everything is "critical", the scale has collapsed and nothing is being separated. Then check the version control answers, which are the other common cause.

What about a spreadsheet nobody uses any more? Assess it once, mark it retired, and delete it — particularly if it holds personal data, as the sample's last row does. An unused file with old pay records in it is a liability with no offsetting benefit at all.

Saving your work

Assessments, settings and the report header are written to this browser's local storage as you type. Export .json is the real save, which Import .json restores anywhere. Export CSV gives you the assessment for spreadsheet work — which is either ironic or entirely appropriate. Reset asks twice, then erases everything. There is no undo.

Accuracy & disclaimer

This is a structured opinion, not a measurement. The score reflects the answers you give, and the weightings are judgements about what usually goes wrong with business spreadsheets — published above in full so you can disagree with them deliberately rather than by accident.

Two cautions matter more than the arithmetic. A high score is a case for looking properly, not for buying software; replacement projects fail more often than spreadsheets do. And a low score does not make a spreadsheet safe — well-controlled files have produced very expensive wrong answers, because the failure is usually a formula nobody checked rather than the tool itself.