Think about that last spreadsheet you circulated to your business partners. Are you absolutely sure it was correct? One hundred percent sure?
How many spreadsheets are found to contain errors? One in ten? Five in ten? According to those in the know at the European Spreadsheet Risks Interest Group (EuSpRIG or "YewSprig"), nine in ten spreadsheets are "materially incorrect". In other words, most spreadsheets contain errors that are potentially of great consequence.
Still have confidence in that spreadsheet? Want to feed your anxiety further? Then read on.
EuSpRIG collates and comments on high profile spreadsheet horror stories that have resulted in the loss of millions of dollars as well as egg on the face of economists at Harvard; forecasters at AstraZeneca; traders at JP Morgan; vote counters at Pueblo County; the CFO of the Town of Framingham; and others. The documented slip ups include copying and pasting errors; data entry errors; logic errors; bad formulas; bad links; accidental publishing of confidential data; confused versioning; hidden rows; and, of course, typos. EuSpRIG offers a number of papers as well as an annual conference promoting best practice but, as a starting point, most of the aforementioned errors could have be avoided if users had received proper training then oversight to ensure their outputs were double-checked and reviewed by peers.
"When we start looking to smartphones and apps, we always have to think about who is being left out, who is not being represented"
Kate Crawford, Researcher, Microsoft
This week's podcast of More or Less: Behind the Stats looks at the pitfalls of big data (Google Flu Trends anyone?) and reminds us not to ignore the old statistical lessons particularly the importance of basing conclusions on representative samples.