Impact of bugs

Economic impact

You may wonder how much impact bugs have on our economy. Trust me, they cost a lot.

In a recent study, researchers of Cambridge University, Undo and RogueWave found that the economic cost of software bugs can be estimated at US$ 312 billion annually. This is more than the GDP of some European countries.

Some high-profile examples come to mind that cost an astounding amount of money.

As far back as 1962, the Mariner-1 space probe mission failed due to a missing hyphen in the source code. This bug caused the guidance system to malfunction, and mission control had to activate the auto-destruct mechanism. Fortunately, the latter functioned correctly, but the incident cost the tax payer US$ 18 million.

More recently, in 2009, a bug in the anti-lock-break software installed in Toyota Lexus cars resulted in the recall of 9 million cars, a financial loss of US$ 3 billion, and, most tragically, the death of four people.

However, not only the effects of bugs come at a price. The time developers spend on finding and fixing bugs is also considerable. The same study quoted earlier also estimates that an average developer spends 50 % of her time debugging.

It is important to catch bugs as early as possible in the development process. A study by the Systems Science Institute (IBM) estimates that the cost of fixing a bug in the quality assurance phase of a project is 15 times that of one found during development. A bug found when the software is already in production is 100 times as expensive.

The damage to a company's reputation as a consequence of defective products can be translated directly into the value of its shares on the stock market as the recent meltdown and spectre bugs illustrated, although long-term consequences may be even more severe.

Impact on science

Although the cost of bugs in scientific software is much harder to determine, it should be clear that they can have a significant impact. Some researchers have expressed concerns about this particular topic, e.g., David Soergel who published an article with the title "Rampant software errors may undermine scientific results". Perhaps Soergel overstates the problem, but it is at least embarrassing when a paper has to be retracted due to bugs in the software used to obtain the results.

A Nature news features discusses leaked emails on climate research in which one of the researchers repeatedly refers to problems in his software as

Yup, my awful programming strikes again.

The article calls for professional training for scientists who develop software packages.

In short, good coding practices and debugging skills are as important to scientists as they are to commercial software developers.