A new system is not a moral duty. For a small firm, the useful question is which workbook is still the right tool, and which one has become the place the business quietly runs. What should stay in the spreadsheet is a shorter list than vendors imply, and a longer list than a transformation slogan allows.
The European Spreadsheet Risk Interest Group keeps the research record on this. Its research page says the majority of spreadsheets contain errors, and it puts that majority above 90 per cent, because people err and because spreadsheets are rarely tested. Ray Panko’s paper for the EuSpRIG 2015 conference, What We Don’t Know About Spreadsheet Errors Today, says intensive inspections found errors in 94% of the spreadsheets studied. The EuSpRIG page also points to earlier research that found material defects in operational models used by large businesses. That work dates from 2007 and 2008. The point is about models that already run a business, not about a list of holiday dates. It is a reason to test the important workbook, not a reason to ban the grid.
Keep the spreadsheet when one person can explain every cell
A spreadsheet should stay when three things are true at once. One named person can open it and explain every number a colleague might challenge. The logic fits on a single sheet a new hire can read in an afternoon. A wrong cell does not move cash, a filing, a price a customer has been promised, or a decision about a person.
That covers a lot of real work. A draft budget you will retype into the accounts. A conference guest list. A one-off comparison of two suppliers. A tracker for a project that will end. These do not need a login, a vendor, and a monthly fee. They need an owner and a habit of checking the total against a source you trust, such as the bank or the invoice.
EuSpRIG’s point about testing is the habit. If nobody has ever changed an input and checked that the output moved the way you expect, you do not yet know the sheet. Do that once before you either keep it or replace it. A sheet that fails the test is not a candidate for a new platform. It is a candidate for a repair.
Move it when the sheet has become the system
Move the work out of the file when any of these is true.
- More than one person edits it, and you cannot say whose version is current.
- It is the only record of money you have collected or are owed.
- A customer, a lender, or a tax office would be entitled to rely on it.
- It decides something about a person: a wage, a bonus, a hire, a benefit.
- The file has links to other files that only one laptop can open.
An operational model is a workbook a business runs on. A five-person firm can still build that kind of model by accident. The invoice tracker that also calculates commission, also emails the customer, and also lives in three inboxes is no longer a list. It is a system without a log. That is the moment to move the ledger into the accounts tool you already pay for, and to leave the spreadsheet as a draft, not as the books.
Fraud sits in the same gap. EuSpRIG notes that code and data live in the same file, which makes a quiet edit hard to see. You do not need a crime story to act on that. If a number can be changed without a name and a date, do not use that file as the record you would defend.
| Keep the spreadsheet | Move the work |
|---|---|
| One owner, one sheet, a wrong cell is only embarrassing | Several editors, and nobody can name the current file |
| A draft you will retype into the real ledger | The file is the only ledger |
| A list that ends when the project ends | A calculation a customer, a lender, or a filing will rely on |
| You have changed an input and checked the output | Nobody has tested it, and it still drives a decision about a person or a payment |
Do not buy software to hide a decision
The broad guides to digital change tend to start from the tool. The firm’s practical guide for SMEs is the wide map. This page is the refusal. If you cannot say what the new system will stop you doing by hand, you are not transforming. You are adding a login beside the file you were afraid to audit.
Supply chains and shop floors have their own version of the same trap. A workbook of deliveries can be the right tool until it is the only record a customer will ask for. The note on practical supply-chain digitisation is about that operating layer. Read it when the sheet has become the promise to a buyer. Do not read it as a reason to replace a holiday roster.
Payments sit in the same split. A sheet of who owes you money is a prompt. The record of what was paid belongs in the ledger. How digital payments help African SMEs with credit is about the rail, not about retiring every grid.
A one-hour inventory
List the workbooks people mailed each other last month. For each, write the owner’s name, whether a wrong cell moves money or a person’s outcome, and whether anyone has tested it. Three piles follow.
Keep, and put the owner’s name in the file name. Repair, because the test failed and the sheet is still the right tool. Move, because it is the system. The move can be dull. Put invoices in the accounting tool. Put the staff list in the place payroll already reads. Leave the analysis sheet as analysis.
If the inventory is mostly “keep”, you do not have a software problem. You have a tidy grid, and that is a good outcome. EuSpRIG’s line on errors above 90 per cent is not a quota of projects. It is a reason the operational pile, the right-hand column, does not get to stay untested because everyone is busy.
Key takeaways
- EuSpRIG’s research page says most spreadsheets contain errors, above 90 per cent. Panko’s EuSpRIG 2015 paper says intensive inspections found errors in 94% of the spreadsheets studied.
- Keep a sheet when one person can explain it, it fits on one sheet, and a wrong cell does not move money or a decision about a person.
- Move it when it is the only ledger, several people edit it, or someone outside the firm will rely on it.
- Test before you replace. A broken sheet moved into new software is still a broken process.
- An inventory that is mostly “keep” is a finished decision, not a failure to modernise.
FAQ
Should a small firm ban spreadsheets?
No. Ban the untested file that is your only ledger or that decides a payment or a person’s outcome. A draft list with one owner can stay.
What does the 90 per cent figure apply to?
EuSpRIG’s research page says the majority of spreadsheets contain errors, above 90 per cent, largely because they are rarely tested. It is not a claim about every holiday roster.
What is an operational model?
A workbook the business runs on: pricing, commissions, a filing, a forecast someone will act on. EuSpRIG’s page points to earlier research that found material defects in operational models.
Is a shared drive a control?
No. If two people can save over each other and you cannot name the current file, the sheet has already become a system. That is a reason to move the record, not to buy a portal first.
What should we do this week?
List the files mailed last month. Mark each keep, repair, or move. Repair means test it. Move means the accounting tool or payroll already has a home for that record.
Sources
- European Spreadsheet Risk Interest Group, “Research and Best Practice”. eusprig.org
- Ray Panko, “What We Don’t Know About Spreadsheet Errors Today”, 2016. arxiv.org/pdf/1602.02601
This article is general information, not investment advice. See section 4 of our Terms & Conditions.
This is the kind of decision we advise on. If you are weighing it for your own business, see how we work or start a conversation.
