Cross-checking is nice, but the research on spreadsheet errors has shown that cross-checks have to look at every single formula and even then will miss about half of all errors if done by an individual. If you ain't doing that kind of cross-checking, you have a whiskey cure. My grandfather used to note that of all things that do not cure the common cold, whiskey is by far the most popular. Makes you feel like you are doing something. You aren't.
Spreadsheet errors look very much like software errors in both frequency and type. Spreadsheet errors are hard to fix in the same way that software errors are hard to fix. Cell protection is one good practice, but spreadsheet development needs good practice across the entire life cycle. Most significantly, spreadsheet developers need to spend about 30% of their time doing testing. The average is closer to 0% now. For research on spreadsheet errors, consider my website, http://panko.cba.hawaii.edu/ssr. I also have a companion website on human error, http://panko.cba.hawaii.edu/HumanErr. The human error website shows that human error is pretty much constant across application domains with comparable complexity and doesn't vary widely across people. Alexander Pope's dictim, "To err is human" is not only true. Today, we can quantify it. And, unfortunately, we can also ignore it.
Slashdot Mirror
Cross-checking is nice, but the research on spreadsheet errors has shown that cross-checks have to look at every single formula and even then will miss about half of all errors if done by an individual. If you ain't doing that kind of cross-checking, you have a whiskey cure. My grandfather used to note that of all things that do not cure the common cold, whiskey is by far the most popular. Makes you feel like you are doing something. You aren't.
Spreadsheet errors look very much like software errors in both frequency and type. Spreadsheet errors are hard to fix in the same way that software errors are hard to fix. Cell protection is one good practice, but spreadsheet development needs good practice across the entire life cycle. Most significantly, spreadsheet developers need to spend about 30% of their time doing testing. The average is closer to 0% now. For research on spreadsheet errors, consider my website, http://panko.cba.hawaii.edu/ssr. I also have a companion website on human error, http://panko.cba.hawaii.edu/HumanErr. The human error website shows that human error is pretty much constant across application domains with comparable complexity and doesn't vary widely across people. Alexander Pope's dictim, "To err is human" is not only true. Today, we can quantify it. And, unfortunately, we can also ignore it.