Often. Researchers who audit spreadsheets that companies actually use find errors in most of them. In a study of 25 spreadsheets at five organisations, 16 contained errors, 117 in total. Many of those errors changed nothing in the result, but the largest shifted a result by more than 100 million dollars. How often things go wrong in the bookkeeping of a small business specifically has not been studied separately.
In short
- Anyone who audits real spreadsheets almost always finds errors. A review study speaks of multiple errors in nearly all large spreadsheets.
- Most errors are small or have no effect. An occasional error is huge, and you cannot see it from the spreadsheet itself.
- The law does not prescribe software, but your records must always clearly show your rights and obligations. Checking against the bank is the simplest protection.
About the research in this article: there is no research on errors in small business bookkeeping kept in spreadsheets. There is research on spreadsheets that organisations use for their work, often with financial calculations. That is the research we use here.
In this article
How often do spreadsheets contain errors?
Most spreadsheets in use contain at least one error. A precise percentage for all spreadsheets cannot be given. Studies measure differently, but they all find errors.
Researcher Raymond Panko summarises the research on spreadsheet errors as follows: nearly all large spreadsheets contain multiple errors, and errors of material size are common. Material here means: large enough to change a decision or an annual figure (Panko, 2006).
Researchers at Dartmouth College previously audited 50 spreadsheets from practice. They found errors in 0.8% to 1.8% of all cells containing a formula, depending on what they counted as an error (Powell, Baker & Lawson, 2009). That seems little. But a spreadsheet with a thousand formulas then contains eight to eighteen wrong cells.
Are those errors also large?
Usually not, sometimes very much so. In a detailed study, about two in five errors had no effect at all on the result. Most of the others were small. A few were huge.
The same researchers audited 25 spreadsheets at five organisations, including two consultancies, a large financial services firm and a manufacturer. They discussed every possible error with the author. Nine spreadsheets contained no errors. In the other 16 they found 117 errors in total (Powell, Baker & Lawson, 2009).
Of those 117 errors, 47 changed nothing in the result. Often a formula pointed to the wrong cell, but that cell happened to have the same value. Of the errors with a measurable effect, most stayed below 10% of the original value. The largest error changed a result by more than 100 million dollars (Powell, Baker & Lawson, 2009).
The researchers also saw large differences between organisations. Some people and organisations produced almost error-free spreadsheets. Elsewhere, bad habits such as typing fixed numbers into formulas were very common (Powell, Baker & Lawson, 2009). So errors are not fate. They depend on how you work.
What goes wrong in practice?
Two types of error keep coming back: a formula that does not include all rows, and a spreadsheet that changes data by itself. Both are invisible if you only look at the result.
A well-known example comes from economics. Two well-known economists published that countries with public debt above 90% of GDP shrank on average, by minus 0.1% per year. Other researchers worked with the original authors' spreadsheet. They found coding errors and omitted data, among other things. After recalculation, average growth was 2.2% per year (Herndon, Ash & Pollin, 2014).
The second example comes from biology. With its default settings, Excel converts some gene names into dates or numbers. An automated scan of leading journals showed that about one fifth of papers with Excel lists of genes contained such errors (Ziemann, Eren & El-Osta, 2016).
The same happens in bookkeeping. A total that does not include the new row at the bottom, an invoice number that Excel reads as a date, a VAT rate typed into a formula. The spreadsheet simply gives a result, without any warning that something is missing.
Why do you not see errors in your own spreadsheet?
Because a spreadsheet has no built-in check. Accounting software works with double-entry bookkeeping: every entry appears on two sides, and they must balance. A spreadsheet calculates whatever you enter, even if it is wrong.
On top of that, the author of a spreadsheet is the worst person to check it. You read what you meant, not what is there. In the research at five organisations, the errors only surfaced when two researchers went through each spreadsheet independently and discussed every point with the author (Powell, Baker & Lawson, 2009).
Organisations themselves rarely have a good system for it either. According to Panko, few organisations have effective controls against spreadsheet errors (Panko, 2006).
May you keep your books in a spreadsheet?
Yes, the law does not prescribe any software. But your records must be set up so that your rights and obligations are always clear from them. The tax authority must also be able to check them within a reasonable time.
This is set out in tax law. Business owners must keep records from which their rights and obligations and the data relevant for tax are clear at all times. They must keep those records for seven years. The records must be set up so that the inspector can check them within a reasonable time (article 52 of the General State Taxes Act).
For a private or public limited company, the Civil Code also applies. The board must keep records from which the company's rights and obligations can always be known (article 2:10 of the Civil Code). If the company goes bankrupt and those records were not in order, the law says the board has performed its duties improperly. Directors then risk becoming personally liable for the debts (article 2:248 of the Civil Code).
How do you limit errors in spreadsheet bookkeeping?
By checking the spreadsheet against something it does not calculate itself. The research shows that errors stay silent until someone looks for them from the outside. Four checks catch most of them.
Reconcile with the bank every month
The closing balance in your spreadsheet must equal the balance on your bank statement. A difference means an error or a missing entry.
Do not put numbers in formulas
Put a VAT rate or tariff in its own cell and refer to it. Fixed numbers in formulas were a common bad habit in the research.
Let someone else review it
Someone who did not build the spreadsheet reads what is there instead of what was meant.
Add up in two ways
Make the sum of the rows and the sum of the columns give the same total. If they do not match, something is missing.
As your business grows, with more invoices, VAT returns and staff, a spreadsheet becomes larger and therefore more prone to error. Accounting software with double-entry bookkeeping and a bank feed removes much of the manual work. It is no guarantee against errors, but it does enforce a balance that a spreadsheet does not have.
Sources
- Panko, R.R. (2006). Spreadsheets and Sarbanes-Oxley: regulations, risks, and control frameworks. Communications of the Association for Information Systems, 17, article 29. doi.org
- Powell, S.G., Baker, K.R. & Lawson, B. (2009). Impact of errors in operational spreadsheets. Decision Support Systems, 47(2), 126-132. doi.org. Figures from the open-access version via arxiv.org.
- Herndon, T., Ash, M. & Pollin, R. (2014). Does high public debt consistently stifle economic growth? A critique of Reinhart and Rogoff. Cambridge Journal of Economics, 38(2), 257-279. doi.org
- Ziemann, M., Eren, Y. & El-Osta, A. (2016). Gene name errors are widespread in the scientific literature. Genome Biology, 17, 177. doi.org
- Algemene wet inzake rijksbelastingen (General State Taxes Act), article 52. Consulted on wetten.overheid.nl, October 2026. wetten.overheid.nl
- Dutch Civil Code Book 2, articles 2:10 and 2:248. Consulted on wetten.overheid.nl, October 2026. wetten.overheid.nl