All Roads Lead To China wondered in several recent posts what proportion of “product failure” is acceptable, and when a recall is justified. What Richard Brubaker calls “product failure” is actually an accident caused by the product (whether it be a design or a manufacturing mistake).

What can Western consumers accept, and what can’t they accept?

One of his examples is a recall caused by 3 baby cribs that caused death, out of 1 million imported pieces. The proportion is .0003%. In these 3 cases, the consumers had “installed the drop-rail side of the crib upside down”, which “creates a gap in the crib that children can slide into and suffocate.” The question is: don’t consumers have “unreasonable expectations” and isn’t this acceptable?

I tend to disagree with this analysis. If it is possible to use the crib in a position that might cause death, then 100% of the pieces are failures, and unacceptable. If only some of them are potentially unsafe, and it comes from manufacturing mistakes, the recall might be too strong a response. But when it comes from poor design by the importer, and ALL the products are unsafe, a recall seems to be justified indeed.

What I call an unreasonable recall

I started really thinking of this issue when reading an excellent book about Intel’s experiences in the 1980s and 1990s. It tells the story behind one of the most expensive recalls of history. It happened in 1994. I am going to quote several excerpts.

Intel found out, too late, a small mistake in the design of their CPUs:

A minor design error on the chip, which caused a rounding error in division once every nine billion times.

An average spreadsheet user would run into the problem once every 27,000 years of spreadsheet use.

When a math professor also found it and the press started covering the issue, Intel did what seemed like the right move:

Our replacement policy was based on our assessment of the problem. People whose use pattern suggested that they might do a lot of divisions got their chips replaced. Other users we tried to reassure by walking them through our studies and our analyses.

However, the press got really nasty, and IBM said they stopped shipping PCs with a Pentium CPU. Then Intel decided to go for a massive recall:

After a number of days of struggling against the tide of public opinion, of dealing with the phone calls and the abusive editorials, it became clear that we had to make a major change.

We decided to replace anybody’s part who wanted it replaced, whether they were doing statistical analysis or playing computer games. This was no minor decision. We had shipped millions of these pieces by now.

Ultimately we took a huge write-off-to the tune of $475 million.

Now THAT’s what I call “going over the top”. I guess it was the price paid for running all these “Pentium Inside” ads and trying to build a consumer brand. Now, if “a rounding error in division once every nine billion times” causes so many consumers to send their chip back to the manufacturer, think of cribs that might kill babies…