Everyone by now, must have heard about the USA being downgraded to AA rating by Standard & Poor’s, but how many know that S&P made a $2 trillion dollar math error when reporting their downgrade to the Treasury Department?
Specifically, S&P projected the gross domestic product instead of projected inflation to calculate what the U.S. government spending and total U.S. Debt would be for the next 10 years. These kinds of math errors aren’t unusual in the fast paced business environment.
Mathematician, Thomas R. Nicely says “The best method of preventing (errors) is to have two or more independent procedures for determining or checking the same result.”
This is done in our analytics work every day. For example, when a download of purchasing history is received for one year from the clients and it’s calculated, it is cross checked with the client’s accounting department’s calculation. If they don’t match, the source is checked to find out the reason.
Another way is to run the numbers (this is called a draft report) by the party to verify your assumptions, calculations and results before the final report is published. This is how S&P uncovered the $2 trillion dollar error. The numbers were run by the Treasury Department and the Congressional Budget Office, before releasing publically, and then adjusted downward the USA’s projected future budget shortfalls once the error was caught.
Lastly, it is found that fact-checking for the “reasonableness’ of the numbers works as well. It’s been the policy for years that if one of our analysts discovers savings for the clients of 50% or more on any SKU it is thrown out or considered as an outlier. The reason being, that it is not realistic or credible in most situations to present the clients with a number that is unlikely to save them money.
All said and done, everyone makes math, spreadsheet and calculation mistakes; however it should be as infrequent as possible. Only by having more than one fact-check, submitting draft reports and inspecting the numbers for reasonableness can one reduce math errors to the lowest number possible.