i-nth logo

We trust spreadsheets more than we should. Because the computer says so, and computers don't make mistakes, we think that the result must be right.

That is, a spreadsheet becomes "truth", losing any uncertainty that it may have had while being developed.

The phenomenon of abstract results gaining a sense of concreteness and reliability, beyond what is reasonable, is called reification.

The problem is partly how spreadsheets work

Spreadsheets are almost certain to have errors. This is partly due to how spreadsheets work:

  • Calculations, and errors, cascade from cell to cell. Spreadsheets are structured as a matrix of connected cells, so an error in one cell is likely to cascade down to another cell, and then to another. Even with a small probability of an error in any cell, it is almost certain that bottom-line results are wrong.
  • Only the surface "presentation layer" of a spreadsheet is immediately visible, while most of the complexity is hidden. A large, well-formatted spreadsheet is seen to be more dependable than a small, poorly formatted spreadsheet. This belief is illogical, as a large spreadsheet is much more likely to have errors due to its size and complexity.

Mostly the problem is us

But spreadsheets have errors mostly because of how humans work:

  • We're human, so we made mistakes. Making mistakes is inherent in how our cognition works. We don't notice most of our mistakes, but they can easily be built into the hidden complexity of a spreadsheet.
  • Spreadsheet complexity leads to cognitive overload. We have limited ability to manage complexity, such as when deciphering the tangled formulae in a spreadsheet.
  • We're overconfident, despite complexity and cognitive overload. It is human nature to be confident about our abilities. Consequently, we tend to not look for errors. If we do happen to notice an error, that simply reinforces our overconfidence in our ability to find errors. We optimistically – and incorrectly – assume that the errors we find are the only errors that exist.
  • The more skilled a user, the less attention they pay to routine processes, ironically increasing the likelihood that errors will go unnoticed.
  • Managers do not recognise spreadsheet risk. By seeing spreadsheets as a simple tool that is easy to use, managers do not devote enough attention to the risks. Like with overconfidence, if we don't look for risks, then we tend to not find them. The perceived absence of risks reinforces our lack of attention.

Most spreadsheet users are surprised to learn that even simple spreadsheets have errors. Even so, it is rare to find a non-trivial spreadsheet that is entirely without errors or significant, unmanaged risks.

When a error is found, people may say that it is a spreadsheet error. However, the problem is not that computers make mistakes, but that people do.