The Cost of Analytical Failure

I spent twelve years on construction projects in the UK and the Middle East. On a site, you learn quickly that the worst moments rarely arrive as surprises. They arrive as a number that looked wrong, a test that nobody liked, and a room that agreed to move on. Deepwater Horizon is the clearest case of that I know, and it opens Chapter 1 of my book, What Are We Missing?

The rig was finishing the Macondo well. The well was in trouble. It had been budgeted at $96.2 million and 51 days, and by 20 April it was almost six weeks behind and more than $58 million over budget.¹ Six days earlier, a BP engineer had written in an e-mail that it had been a "nightmare well".¹

The first decision: skip the test of the cement. On the morning of the accident, the BP team judged the cement job a success. A crew from Schlumberger was on the rig, ready to test whether the cement had really sealed the well. The team sent them home on the 11.00 a.m. helicopter, "saving time and the $128,000 fee". The reason given was that everyone on the rig was "completely satisfied" with the job.² Satisfaction is a feeling. The test would have been a measurement.

The second decision: explain away the reading. That evening the crew ran a negative-pressure test, which checks whether the well holds when pressure is taken off it. The drill pipe was bled down to zero three times. Each time, the pressure climbed back, to 1,400 psi. Someone suggested a "bladder effect". One of the men in the room later said he did not recall anyone agreeing or disagreeing.³

The crew then ran the test again, this time on a different pipe, the kill line. It showed no flow for 30 minutes. So there were two readings from the same well. One said safe, one said not. The National Commission found that the two readings were never reconciled, and that the 1,400 psi on the drill pipe "could only have been caused by a leak into the well". At 8 p.m. the well site leaders still concluded the test had confirmed the well's integrity.³

That is the cost of analytical failure in one sentence. They had the contradiction in front of them and went with the reading that let them carry on.

The third failure: a warning that was missed. At about 9.01 p.m., with the pumps running at a constant rate, the drill-pipe pressure began to rise. Over seven minutes it moved from 1,250 to 1,350 psi.⁴ That is a well telling you something is coming up it. The explosion was at 9.49 p.m. Eleven men were killed.⁶

None of this is about people being stupid. The men on that rig were experienced and many of them acted bravely that night. The problem was method. A test was skipped to save time. A contradiction was resolved by the more comfortable answer. A rising pressure was not read as a warning. Each step looked reasonable and each step assumed the one before it was sound.

The consequences ran for years. The well was not capped until 15 July, 87 days after the explosion.⁵ BP later pleaded guilty to 14 criminal counts, including 11 counts of felony manslaughter, and admitted that its two senior men on the rig had seen clear indications that the well was not secure.⁶

What this means for the rest of us. You will probably never run a drilling rig. But you will meet a version of the same moment. A report that does not quite add up. A check that costs time. A number that you could explain away if you wanted to. The question that matters is the one the Macondo crew did not ask in time: what are we missing?

That is what Chapter 1 of What Are We Missing? is built on. It takes the night apart hour by hour, and then gives you a method, called PRISM, for catching the same failure in your own decisions. The book cites 498 works, with every claim traced to its source. It is planned for 13 February 2027. The sources are listed in full, and free to download. Check them. That is what they are for.