Boeing 737 MAX, feedback loops and incentives
A reading of Flying Blind by Peter Robison - "A culture, if you can keep it"
Despite what you might think from the last few posts, I’m interested in things other than just Ottawa. I have just finished reading Flying Blind - The 737 MAX Tragedy and the Fall of Boeing by Peter Robison. As the title suggests, this is a book about what went wrong with the Boeing 737 MAX. This is the plane that was grounded in 2019 after two disasters that were determined to be the fault of a cover-up at Boeing and poor oversight from the FAA.
My interest in this book isn’t so much on the actual airplanes or the disasters, but on organizational culture, design, incentives, etc. Corporations like Boeing are obviously complex systems. They are first and foremost made up of people with their own beliefs and incentives. I wanted to draw some lessons from Boeing on how to properly design organizations and avoid disaster.
Boeing and its transformation
The first half of the book is a story of Boeing itself and its origins during World War I building wooden planes. Robison makes it clear that Boeing was a corporation run by engineers. Their slogan was “Working Together”. They built relationships with their stakeholders - FAA, airlines, suppliers, etc. Starting in the 1990s after a merger with McDonnell Douglas, Boeing became a corporation run by financiers off the model of General Electric under Jack Welch. Their slogan became “More For Less”. They embarked on cost cutting, nickle and diming suppliers, litigious, outsourcing, union busting, etc. Boeing gave up its focus on long-term thinking for short-term profit.
The Dreamliner was a sign that something was rotten in Boeing. Robison describes a number of shifts in design and development work on the Dreamliner compared to previous planes. Design work was often outsourced piecemeal to subcontractors rather than designed in house. The sticker price would often be much cheaper per part, but it led to increased overhead work (e.g. when designing in house, specs could be less detailed since they could easily change if necessary. Outsourced work had to be done to the contract or lawyers would get involved from either side, so specs were much more detailed. More detail, more cost to produce). This is often the case for other large projects. I wonder if we can draw larger conclusions from this on why building things in North America is so expensive now.
Talent often works in cycles. Smart people want to work with other smart people on hard problems. Boeing would attract smart people, which would improve perceptions of Boeing, which would attract more smart people. Virtuous cycle! Once the culture of Boeing shifted, layoffs happened, fighting with unions, outsourcing design and production work, etc the cycle shifted. Morale was hurt, so people would leave, which hurts morale more so more people leave, etc. Robison makes it clear that there were many smart engineers who had worked at Boeing, and by the time the 737 MAX was nosediving into the earth, were no longer there.
Squeezing suppliers. They would ask for cuts of 15%, then turn around and ask for a further 10% cost. Maybe you get the cost savings, but you start ruining your relationship with suppliers. Byrne Hobart wrote about Minsky Moments in supply chains where suppliers will be willing to put up with a lot, until they aren’t1. Pissing off suppliers works as a vicious cycle as well.
The 737 MAX
So what went wrong with the 737 MAX? As with most things, it wasn’t one thing, it was a series of unfortunate decisions and culture changes over time that laid the groundwork for the disasters.
The air certification process with the FAA is expensive, involving lots of engineering time, lawyers, following regulations, etc. The 737 MAX started as an update to the 737. An updated model doesn’t need to necessarily go through the extensive new certification process, so Boeing pushed hard - legally, engineering-wise, etc. Customers were on board with this. New planes mean extensive pilot training. Airlines need to pay for pilots to go to training, and that’s expensive, so airlines that were promised no significant training costs were keen on keeping it that way as well. Boeing essentially ran the FAA - decision making in the FAA became subservient to Boeing based on political changes.
Breakdown in feedback loops
At the beginning, Boeing would embed staff and trainers with airlines (something Airbus still does). These people would get feedback from the airlines and actually see what the pilots are doing. This would influence cockpit design. This broke down when training was outsourced.
Boeing also, because the 737 MAX was not a new plane, justified fewer test flight hours. Less flight hours, less chance of finding things that could go wrong.
MCAS
A change in the 737 MAX from 737 made the plane unstable under certain conditions. Because of the relentless cost cutting and pace of development, a fix was needed. Based on the incentives, a software fix was deemed best, specifically MCAS. This was a lazy change, because it didn’t fix the fundamental issue with the plane, it just hid it in software. It would adjust the horizontal stabilizers itself, and wouldn’t allow the pilots to override it, which is not something planes generally do.
Under certain conditions, this led to pilots fighting the plane didn’t work properly. Combined with the breakdown in feedback loops, this became deadly. As well, because Boeing was pushing for not having to redo the certification process, this behaviour was undocumented. And because the 737 MAX was no different to the 737 (or so Boeing said), pilots only had to have a few hours of iPad based training, not actual simulator training. So pilots had to fight undocumented behaviour, with no actual training.
Nickle and diming
It actually goes even a little bit further than this as well. The two crashes occurred on airlines that bought a lower trim package of the 737 MAX, that did not have certain sensors on board. Because they didn’t have those sensors, MCAS relied on a single source of data. It appears that on both the Lion Air and the Ethiopian Airlines flight, something went wrong with the single sensor. The sensor provided faulty data to MCAS, which went into overdrive. And since the behaviour was undocumented and poorly tested, this wasn’t expected by either Boeing, the FAA, or the airlines.
Summary
So how did the 737 MAX debacle happen? None of the above problems individually caused the problems, but they all contributed and were necessary for it to happen. Robison makes clear that this wasn’t all fradulent (although Boeing paid huge fines). A lot of it, in terms of hiding features, were accidental and the pieces were never really put together properly. To generalize:
Change in thinking from long-term to short term (change in organizational values from engineering and safety to cost cutting)
Breakdown in feedback loops, throwing away a culture of learning
Incentives aligned against safety and towards cost cutting
Regulatory capture of the FAA
What takeaways can we get from this? One thing that’s apparent throughout this book is that building airplanes is hard work. The complicated manufacturing process probably is more susceptible to failing, especially when it has to be near perfect every time. Defects from the widget factory are returned and scraped. Defects from Boeing kill people. This was also compounded by airplanes getting more and more complicated and filled with software.
Culture is also hard to build. Boeing’s culture was built over decades, and ruined over a smaller amount of time.
Institutions, like the FAA, are even harder to maintain. The experts of aviation safety work for Boeing because they get paid more there. A section of the book goes into FAA’s salary structure as well, which apparently is partially related to preventing delays in certification. That’s just bizarre, unless the purpose is to rubberstamp everything Boeing does (which was the point).
How can we align institutional incentives towards goals? How can we design an institution that achieves our goals over a long horizon? Safey isn’t the concern of my organization, but quality is. How do we design an organization that favours long-term thinking over short-term profit? I think talent is an important part of that. Hiring the correct people, and ensuring the talent feedback loop stays virtuous instead of decaying. Having the correct people at the top (“the tone at the top”) also seems very important.
There’s probably some lessons in the dangers of mergers as well. Boeing absorbed McDonnell Douglas and ended up turning into something else. There’s always M&A occurring, and maybe some of these companies aren’t thinking about the long-term effects on their own culture by absorbing employees who haven’t already been subsubed into it.
Further Reading
Flying Blind - The nook in question
Boyd: The Fighter Pilot Who Changed the Art of War - I wrote about this before, but Boyd was intimately involved in fighter plane design and there’s a lot of the same lessons from this book.
A Minsky moment is a sudden, major collapse of asset values which marks the end of the growth phase of a cycle in credit markets or business activity. - Wikipedia
You might find this course interesting: Unethical Decision Making in Organizations - University of Lausanne. You can take it online for free with Coursera. It looks at these type of failures and what went wrong in the human decision making and what effect from corporate culture.