Reports of Spreadsheet Errors in Practice

Ever since the earliest days of spreadsheeting, there have been reported incidents of errors. On the one hand, these reported errors may be only the tip of a very large iceberg. On the other hand, the vast majority of spreadsheets may be correct, so that these few reported incidents may be very misleading. However, data from both field audits and laboratory experiments in spreadsheet development indicate that spreadsheet errors are fairly common.


See more examples at the EuSpRIG website.


Business Week "How Personal Computers Can Trip Up Executives," (2861) September 24, 1984, pp. 94-102 passim.

Forecast sales of $55 million for planned new computer product. Forgot planned price discount on key components to one supplier. Forecast was $8 million over. It was caught in time to prevent damage.
End users shared diskettes containing old data. In one case, a person ordered 30,000 units of a $4 part; but the plan had changed and the company only needed 1,500.
A Midwestern firm's estimated taxes were $5,000 off from a correct paper and pencil value. The spreadsheet had an incorrect formula for assessing salvage value.
Two spreadsheets with 15,000 cells were used to project the market for CAD equipment. Numbers were rounded off to whole dollars. But the inflation multiplier, which should have been 1.06, became 1. Without inflation, the market was underestimated by $36 million.

Casimir, Rommert J., "Real Programmers Don't Use Spreadsheets," ACM SIGPLAN Notices (27:6) June 1993.
Argues that spreadsheet programs should not be taken seriously because they cannot solve the Towers of Hanoi problem easily. [ed.: Actually, with macro languages and Visual Basic for Applications, such arguments are no longer correct.]

Crawford, M.D. "Re: Microsoft and Lotus Spreadsheet Errors," The Risks Digest (16:60) http://catless.ncl.ac.uk/Risks, December 5, 1994.
Describes an error in a version of Excel for the Macintosh in which values in linked spreadsheets are incorrectly calculated.

Creeth, R. "Microcomputer Spreadsheets: Their Uses and Abuses," Journal of Accountancy (159:6) June 1985, pp. 90-93.
According to industry experts, 1/3 of spreadsheets have errors.

Dhebar, A. "Managing the Quality of Quantitative Analysis," Sloan Management Review (34:2) Winter 1993, pp. 69-75.
Fortune 500 firm uses discounted cash flow to evaluate investment proposals. The formula and discount rate were established long ago, were never documented, and were made by a person who had left the company. Although the prime rate rose from 8% to over 20% between 1973 and 1981, the spreadsheet was kept at 8%.
In a consolidation for a large multinational firm, divisions made their own assumptions about inflation rates, interest rates, and economic growth.
Stated that the only way to be sure is to do step-by-step debugging.

Ditlea, S. "Spreadsheets Can be Hazardous to Your Health," Personal Computing (11:1) January 1987, pp. 60-69.
A Florida construction company used @sum to total numbers in a range. The range was not changed when an item was added, so the added item was not summed. This caused the company to underbid the project by a quarter of a million dollars. The company sued Lotus.
Larry Nipon, a consultant, found an error that would have cost $1.5 million had it gone unchecked. The spreadsheet was 85 cells wide. The cell reference designation was incorrect. It looked like a great year, but it seemed to good to be true. The error was caught with Cambridge Spreadsheet Analyst, a spreadsheet auditing program.
Eric Klasson, a Houston consultant with Price Waterhouse audited 4 spreadsheet models. He found 128 errors covering 120 line items. Some formulas were applied differently to two subsidiaries. The spreadsheets had already been in use for months.
According to Input Research, 1/3 of all spreadsheets have errors.

Freeman, R.M. "A Slip of the Chip on Computer Spread Sheets can Cost Millions," The Wall Street Journal, August 4, 1984.
A Dallas Oil and Gas company's spreadsheet error resulted in millions of dollars being lost. Several executives were fired.

Gilman, H. & Bulkeley, W. "Can Software Firms be Held Responsible When a Program Makes a Costly Error?" Wall Street Journal (CCVII:24) August 4, 1986, p. 17.
More on the Florida construction company problem.

Godfrey, K., "Computing Error at Fidelity's Magellan Fund," The Risks Digest (16:72) http://catless.ncl.ac.uk/Risks, January 6, 1995.
Describes how a multibillion dollar reporting error occurred at Fidelity Investments. See Savitz.

Hayen, R.L. & Peters, R.M. "How to Ensure Spreadsheet Integrity," Management Accounting (60:9) April 1989, pp. 30-33.
A Florida construction company used @sum to sum a range. The range was not changed when an item was added, so the added item was not summed. The company sued Lotus. Data from Gillman and Bulkeley.
A Dallas Oil and Gas company's spreadsheet errors resulted in millions of dollars being lost. Several executives were fired.

Savitz, E.J., "Magellan Loses its Compass," Barron's (84:50) December 12, 1994.
At Fidelity, a spreadsheet was used to report distributions for various funds. For the huge Magellan fund, a $4.32 per share capital gains distributions was forecast in November, and investors were notified. However in December the company announced there would be no distribution. A clerical worker put the wrong sign in front of a $1.2 billion ledger entry. This "created" a $2.3 billion gain in place of the real $0.1 billion loss. This may have affected buyers, some of whom may have sold to avoid the distribution and missed a price rise, others of whom may have waited to buy to avoid the distribution and also missed the price rise.

Schofield, J., "Beware of Spreadsheets," Management Today, February, 1992, pp. 39-40. Cited in Cragg & King, 1993.

Simkin, M.G. "Micros in Accounting: How to Validate Spreadsheets," Journal of Accountancy (164:5) November 1987, pp. 130-138 passim.
Describes several cases of errors.

Slade, R. "Is it Getting too Easy?" The Risks Digest (13:45) http://catless.ncl.ac.uk/Risks, April 28, 1992.
In an error made at a Microsoft technical seminar, the presenter copied cells to get a growth rate of 10% from year to year. This method accidentally produced a constant dollar increase, not a constant percentage increase. Fourth column was $1300 instead of the desired $1331.

Woodbury, G. G. "Re: 'Computer Error' in Durham N.C. Election Results," The Risks Digest (9:42) http://catless.ncl.ac.uk/Risks, November 13, 1989.
In a North Carolina election, results of an election were about to be incorrectly posted. Mr. Woodbury, using a calculator, detected an inconsistency. Examination found an incorrect cross-tabulation in the spreadsheet being used to post the results.

Copyright Panko 1997-2005.