Spreadsheet users need to recognize that spreadsheet error is a real and substantial risk.
This is difficult to do, as we are generally overconfident about the quality of our spreadsheets, believing that they are much better than they really are.
One way to help users recognize spreadsheet risk is to have them test their own spreadsheets. They are often shocked by what they find.
We make way more errors than we recognize
In experiments where experienced people built simple spreadsheets, the people working alone thought, on average, that there was an 18% probability that they made an error. In fact, 86% of them made errors. People also thought that they did better than others, with no basis for making such an assertion.
These results are consistent with general human behaviour across essentially everything we do. A famous survey reported that more than 80% of vehicle drivers say they are more skilful and less risky than other drivers. A large meta-study of many different domains confirmed that there is a "tendency for people to perceive their abilities, attributes, and personality traits as superior compared with their average peer". In other words, we tend to be systematically overconfident.
Training can help raise awareness
An effective way to reduce overconfidence, and manage risk, is through better training. Most spreadsheet training is about the software features – for example, learning the steps for constructing a PivotTable. While such training is necessary, it is not sufficient for constructing good spreadsheets. Users need to go beyond learning about the software features, by also learning about spreadsheet risk and overconfidence.
Once spreadsheet users are aware of the risks and their overconfidence, they can focus on learning good practices to manage the risks. Users who understand the risks tend to produce better spreadsheets.