2. Spreadsheets are widely used on core corporate financial modeling, creating the need to control error rates Source: Panko, R., “What we know about spreadsheet errors”; Hall, M., “A Risk and Control Oriented Study of the Practices of Spreadsheet Application Developers”; CODA; Softtrax.com; Most firms use spreadsheets for financial reporting… … with a high prevalence on budgeting and forecasting… … and in highly relevant situations Share of U.S. firms that use spreadsheets for financial reporting Spreadsheet usage with 118 US business leaders, 2004 Relevance level of spreadsheet corporate use in Australia, 1996 100% = Australian firms 100% = U.S. firms 100% = U.S. firms No spreadsheet usage (5%) No spreadsheet usage in budgeting and forecasting (15%) Not Important (7%) Important (54%) Spreadsheet usage (95%) Spreadsheets used for budgeting and forecasting (85%) Critical (37%)
3. Astounding error rates, that increase with sheet complexity, contrast with poor error detection rates Source: Panko, R., “What we know about spreadsheet errors”; Galleta, et. all Complex spreadsheets are about 100% likely to contain errors Error detection rates in debugging tasks is the rate of detected errors by novice CPA and MBA students Is the average cell error rate in American corporations 2-6% 54% Is the rate of detected errors by expert CPA and MBA students of 116 audited corporate spreadsheets have errors 88% 57% is the rate of detected errors by undergrad students of corporate sheets with over 150k lines have errors 91% 63% audited spreadsheets with over 2,300 rows or 2,200 formulas have errors is the rate of detected errors by undergrads in groups of 3 students 100% 83%
4. Logic, quantitative and software are the three major and most frequent types of spreadsheet modeling errors Source: Raffensperger, J. “The Art of Spreadsheet” Logic errors The model is well written, but the understanding of the modeled situation is incorrect Wrong thinking Values typed as inputs are wrong or wrongly typed, generating erroneous outputs Wrong inputs Quantitative errors Formulas are wrong, point to wrong cells, point to label rather than number or point to a blank Accidental logic Formulas are accidentally overwritten with constants or copy/pasted as values rather than formulas Formula overwriting Software errors Different spreadsheet software may behave differently (e.g., Microsoft Excel and Lotus 1-2-3) and change with versions Unexpected behavior Spreadsheets have weaknesses (error prone, little scalability and hard automation) and may not always be the best tool for modeling Spreadsheet fit
5. There are, however, several approaches to prevent spreadsheet modeling errors * With different modelers and possibly different modeling tools/software Source: Raffensperger, J. “The Art of Spreadsheet” Clear design and formatting enables easier reading, interpreting and debugging Ensuring that “apples vs. oranges” situations do not occur in the model is paramount Good design Units of measure Reality check with different outputs to test validity of outputs and functionality of model Having a second person checking the model for correctness improves error detection rates Reality checks Double check Comparing model outputs with known situations to assess correctness of model Developing models in parallel in an organization* leads to lower error rates Known outputs Shadow modeling
6. In conclusion, it is important to control modeling error rates to mitigate the negative effect on business decisions Source: Panko, R., “What we know about spreadsheet errors”; Galleta, et. all