temperatures and O-ring failures. These concerns became more salient when
a cold front and temperatures of 30 degrees Fahrenheit were forecast for
January 28. Because of the cold, a prominent Morton Thiokol engineer rec-
ommended against the launch. But this advice was overruled, and the warning
was not communicated to top NASA officials. The temperature was 36 degrees
Fahrenheit at launch time.
The following temperature line shows the data that NASA scientists gath-
ered on the eve of the launch. The diagram shows the launch temperatures for558 Chapter 13 The Value of InformationO-Ring Wear No O-Ring Wear Total
Temperature65° 31720
Temperature65° 404
Total 71724The message of the table should be clear. Once we have separated out the most
frequent outcome, high-temperature launches showing no wear, we are left
with four launches made at lower temperatures (below 65), all of which experi-
enced O-ring damage.Thus, the table offers strong evidence that low-temperature
launches entail O-ring risks.
By failing to incorporate the results of the “uneventful” flights in addition
to the “problem” flights, NASA came up short in its risk assessment. To put this
as simply as possible, data in all four cells of a two-by-two table are needed to
establish an association between any two factors. Why did NASA miss such an5053 5758 63 70 75
the previous seven shuttle flights that had experienced O-ring wear. These “prob-
lem” launches occurred at both high and low temperatures, so the diagram
reveals no discernible association between low launch temperatures and O-ring
wear. Five of the seven flights had shown wear on a single O-ring, but the launch
at 53 degrees Fahrenheit had experienced wear on three rings. An indication of
low-temperature problems, perhaps? However, the launch at 75 degrees also
exhibited abnormal wear, this time on two O-rings. It is not surprising that NASA
scientists saw no real O-ring risks from low-temperature launches.
Was there evidence available that NASA overlooked? The answer is yes.
NASA’s crucial error was its failure to appreciate the evidence of the 17
“uneventful” flights. In fact, every one of the launches that were free of O-ring
damage occurred at temperatures of 65 degrees and above. Against this back-
ground, the incidence of O-ring wear at relatively low temperature looks quite
damning. For instance, it is instructive to list the outcomes of all 24 launches
in a simple two-by-two table.c13TheValueofInformation.qxd 9/26/11 11:02 AM Page 558