When One Gauge Disagrees
On Deepwater Horizon two pressure readings disagreed, and a familiar explanation from a respected hand made the disagreement go away. Familiar, senior and convenient are not evidence.
On the afternoon of 20 April 2010 the crew of the Deepwater Horizon drilling rig ran a test on the well they had just finished. It is called a negative-pressure test. The idea is to take pressure off the well and see whether it holds. If pressure builds back up, the well is not sealed.
According to the report of the national commission set up to investigate the disaster, the test began at about 5 pm. After the pressure was bled off, it repeatedly built back up. A second test followed. This time pressure on one line, called the kill line, went to zero. On the drill pipe it did not. The pipe kept showing elevated pressure.
The commission reports that, according to BP witnesses, a senior rig hand said he had seen this before and explained the reading away as a "bladder effect". The men in the control shack then decided that no flow from the open kill line meant the test had passed. A toolpusher who had been present earlier was described by a colleague as convinced that something was not right, but his shift was over.
The well blew out at about 9:50 that night. Eleven people died.
I want to be careful here. The investigations did not agree on every detail, and one scientific review noted it may never be possible to establish exactly how the failure happened, because the people who could say were among the eleven who died. I am not claiming that one explanation or one person caused it.
What I take from it is narrower. There were two readings that disagreed, and one explanation arrived that made the disagreement go away. The explanation was familiar, it came from someone respected, and it let the work move on. Those are three good reasons to believe it and none of them is evidence.
It helps to ask a plain question whenever two measurements disagree: what would I expect to see if the comfortable explanation were wrong? If the answer is "the same thing I am seeing", then the explanation cannot be tested, and it should not be relied on. Better to run the test again a different way than to accept the story that arrives first.
It is the case the book opens with. Not because the story is unusual, but because the pattern is not.