Why Good Companies Break Under Pressure

Capable organizations rarely fail from weakness. They fail when pressure gets misread, and the break runs along the grain of their strengths.

Early in my career I ran a program building a ruggedized field radio for a defense customer. We had committed to delivering prototypes nine months from kickoff, and that number was not a guess. We modeled the schedule, ran Monte Carlo analysis to size the risk sitting in each path, and argued about the output. Then we priced the work as a fixed bid, which meant the date was also the economics. If we slipped, we would not be paid more. We would spend more of our own money finishing.

So the date carried three things at once: a commitment to a customer, the margin on the program, and the thing I valued most in how we worked, which was that we did what we said we would do. That was the firm’s reputation and it was my own standard. I believed then, and I still believe, that it was most of the reason a client would choose us over someone cheaper.

Then a senior engineer named Dan found me in Amsterdam.

We were in transit to a customer site in southern England, about to run a global program review with vendors dialing in from several time zones. Dan told me the software was in trouble. The radio ran on an off-the-shelf processor module, and getting the operating system onto it meant working through a four thousand page manual to understand what the part would and would not do. We had underestimated it, and not by a margin anyone could absorb inside the plan. He needed three more months.

I said no. I walked him through the commitment, the analysis behind the nine months, and the fixed bid, and I told him the date was not one of the variables available to us. Then we walked into the program review and did not raise it, because I had already decided there was nothing to raise.

We shipped three months late.

Dan was not just right. He was right to the month, on the timeline he handed me in Amsterdam, which means the information the program needed had arrived exactly on schedule and I was the part of the system that stopped it.

That much I worked out reasonably fast, and it is the smaller of the two failures. The larger one took years, because it did not announce itself as a missed date. It announced itself as better-looking dates.

That conversation was just the two of us, but the team learned from what happened afterward. A senior engineer had brought real information to the person running the program and been told the answer was already fixed. I did not intend to teach anything in that exchange. I taught something anyway, and it was more precise than “Mark is unreasonable.” What the team learned was that raising a risk near a commitment was expensive, and that the expense landed on the person who raised it.

After that, risk stopped arriving as conversation. It started arriving inside the estimates.

The numbers got longer. Then they got longer again. This was not dishonesty, and being careful about that distinction matters, because reading it as dishonesty is exactly how a leader fails to see it. If the date cannot move once it is set, and the cost of saying so falls on you, then the only place left to put uncertainty is the number you give me before the date is set. So that is where people put it.

The one variable a person could still move on their own was the estimate, so that is the variable they moved.

The second-order effect turned out to be worse than the padding. When estimates carry buffer nobody names, committing to one stops meaning what it used to mean. Accountability loosens, because accountability had come to mean volunteering for something that would eventually hurt. Scope conversations got softer. Fewer people wanted their name on a date. An organization built around doing what it said it would do was becoming an organization that was careful about what it said it would do, and those are not the same thing, though they look identical on a status report.

Nobody on that team lacked skill or commitment. They read what was actually valued and responded to it accurately. What I had shown them was valued was the appearance of a held commitment, so that is what they produced.

The failure was mine, and it was specifically a reading failure. Real pressure arrived, I translated all of it into a single internal demand, and I passed that demand down while keeping the reasoning in my own head, where it helped no one.

So let me state the position plainly, because it is not a story about one program:

Good companies do not usually break because pressure exceeds their strength. They break because pressure enters the organization carrying real information, gets translated into internal demands that no longer carry that information, and then gets routed through whatever the company does best.

The strength executes. The signal is gone. What breaks is not the weakest part of the company but the part that was working hardest.

What it looks like at scale

That claim needs evidence beyond my own bad program, because on its face it sounds like a consultant’s paradox. The most instructive corporate breakdowns of the last two decades have something rare: thorough, public, independent investigations that let us watch what the pressure actually did on its way through the building.

For two decades, Wells Fargo was the bank other banks were told to emulate, and the center of its reputation was cross-selling: deepening each customer relationship by adding products, checking to savings, savings to credit card, credit card to mortgage. The ambition had a slogan, “going for gr-eight,” eight products per household, and the company’s ability to execute on it was treated by analysts as evidence of a superior sales culture.^[1]

Then, in September 2016, regulators announced $185 million in fines because employees had opened millions of accounts customers never asked for. The bank’s own later review put the number of potentially unauthorized accounts at roughly 3.5 million.^[2] Thousands of employees had been fired over sales integrity issues across the preceding years, which means the behavior was not hidden from the company. It was being processed by the company, as a disciplinary matter, one employee at a time.

Now look at where the pressure came from and what happened to it in transit. The external signal was real: retail banking margins were compressing, and deepening existing relationships was a legitimate strategic answer. At the top of the company, that is what the pressure meant. But by the time it reached a branch in Phoenix or Newark, it had become a daily sales quota, tracked on scorecards, read out on conference calls, with district managers ringing branches multiple times a day for numbers. The board’s investigators were precise about this: the problem was not that goals existed but that the goals were, in their words, unattainable in many markets, and that the culture treated missing them as a personal failing rather than as information about the goal.^[3]

A teller opening a phantom savings account was not confused about ethics. She was responding accurately to the pressure that actually reached her, which was no longer “deepen customer relationships” but “produce eight by Friday or explain yourself.” The original signal had been stripped somewhere around the regional level. What remained was demand with no meaning attached, and demand with no meaning attached gets satisfied by whatever means the local environment permits.

That is the same move my engineers made, at a scale with no fines attached. She could not produce eight real relationships, so she produced eight accounts. They could not safely commit to a real date, so they produced a safe one. In both cases the measure was satisfied by moving the input rather than the substance, which is what people do when the substance is out of reach and the measure is not.

And notice the channel it ran through. Wells Fargo’s sales machine was not a flaw that ethics training failed to catch. It was the company’s crown jewel, the most developed, most measured, most celebrated capability in the building. The pressure did not find the weak spot. It found the strong one, because the strong one is what an organization reaches for when it is told to produce.

The airplane the schedule built

The Boeing case runs on the same mechanism at higher stakes.

In late 2010, Airbus launched the A320neo, a re-engined version of its bestselling narrow-body, and began taking orders at a pace that threatened the 737’s franchise, including at American Airlines, a customer Boeing had held for decades. Boeing faced a genuine strategic fork: spend roughly a decade designing a clean-sheet aircraft, or re-engine the 737 and get to market fast. It chose the 737 MAX, and with that choice came a constraint that shaped everything downstream: the new plane needed to fly like the old one, so that airlines would not need to put pilots through expensive simulator retraining.^[4]

The larger engines changed the aircraft’s aerodynamics in certain conditions, and the engineering answer was MCAS, software that would adjust the plane’s pitch behavior so it handled like its predecessor. Along the way, MCAS’s authority was expanded, its reliance on a single angle-of-attack sensor was retained, and references to it were minimized in materials pilots would see, all in service of the governing constraint: no new type rating, no schedule slip. The House investigation documented the internal weather this produced, including a program culture employees described as schedule-driven, countdown clocks in meeting rooms, and engineers reporting pressure not to raise issues that could slow certification.^[5] Two crashes, Lion Air Flight 610 in October 2018 and Ethiopian Airlines Flight 302 in March 2019, killed 346 people and grounded the fleet worldwide.

It matters that Boeing was, by any reasonable account, one of the finest engineering organizations in the world. The people who built the MAX were not careless, and the investigations do not describe carelessness. They describe an organization faithfully executing the constraint that had been handed down, long after the constraint had detached from the signal that justified it. The external pressure said: Airbus is taking the market, respond or lose it. By the time it reached the engineering floors, it said: the schedule and the no-retraining commitment are fixed, fit everything else around them. Engineers optimized brilliantly inside a frame nobody at their level had the standing to question. The excellence was real. It was pointed at the wrong invariant.

The same signature appears in the General Motors ignition switch record, where the Valukas investigation found engineers categorizing a stalling defect as a customer convenience issue for years, inside a culture famous for cost discipline and for meetings where everyone nodded and nothing moved.^[6] Different industry, same translation failure.

The pattern has a longer record

Danny Miller documented this trajectory across dozens of firms three decades ago and called it the Icarus paradox: outstanding companies extend their defining strength until it becomes the instrument of their failure.^[7] What Miller framed as strategic momentum, the investigation records let us see at ground level, as thousands of individually reasonable responses to pressure that no longer said what it meant.

This is why the break runs along the grain of the strength. My radio program and Boeing’s are the same shape at wildly different scale, and the difference in outcome is a difference in stakes rather than a difference in mechanism.

The mechanism, stated plainly

Strip the details and the pattern repeats in three segments.

First, the pressure is real and external, and at the top of the organization it is legible. Wells Fargo’s leadership could read margin compression. I could read a fixed-bid contract and a customer commitment. The failure does not begin with a false signal. It begins with a true one.

Second, the signal gets converted into an internal proxy, and the proxy travels down while the reasoning stays at the top. A quota. A schedule. A fixed constraint. Each layer passes on the demand and, usually without intending to, strips a little more of the meaning. Three or four levels down, people are no longer responding to the market or the contract. They are responding to the artifact, and the artifact has stopped being information and become weather.

Third, the organization routes the demand through its most developed capability, because that is what capabilities are for. A sales culture sells its way to the number. An engineering culture engineers its way to the deadline. A cost culture categorizes its way around the expense.

A capable organization under misread pressure does not stop executing. It executes the misreading.

What it costs, specifically

This is where the argument usually gets left as a cautionary tale, and it should not be, because the costs are mechanical and they land in places a leadership team is already watching.

• Decision quality falls while decision speed rises, which is why it stays hidden. A proxy is easier to decide against than a signal. “Does this protect the date” resolves in seconds; “does this serve the reason we set the date” takes a conversation. The organization gets faster and more decisive at exactly the moment it is deciding against the wrong thing, and on any dashboard that reads as improved execution.
• Every downstream decision gets made against corrupted inputs. Once estimates carry buffer nobody names, they stop being forecasts and become negotiating positions. Work gets sized out of existence because a padded number made it look too expensive. Investment goes to the wrong place. The organization is still deciding carefully, using numbers that no longer describe reality.
• Cross-functional throughput degrades by never being invoked. When pressure runs through the strongest capability, the parts of a response that would have required another group are not refused. They are never proposed. On the radio program, nobody suggested renegotiating scope with the customer, because the problem had been framed as a delivery problem and delivery was ours.
• Prioritization starts happening against the artifact. Once the proxy is in circulation, every subsequent tradeoff gets adjudicated against it rather than against the thing it was standing in for. A quarter later you have a coherent, well-argued, disciplined set of priorities, all of them derived from a number that lost its meaning three levels up.
• The price of carrying bad news rises exactly when the news matters most. In every case above, including mine, the information needed to prevent the failure existed inside the organization long before it became a problem. What failed was the path that information needed to travel. Trust determines whether someone spends their own standing to tell you something you do not want to hear, and that spending is discretionary. It stops without an announcement, and output holds for a while after it stops.

The radio program did not break the firm. We delivered three months late, the client stayed, and the loss stayed inside our own margin, which is a small enough bill that I could file it under a hard program and move on. Most organizations do. That is the part worth sitting up for: this mechanism produces the survivable version far more often than the catastrophic one, and the survivable version is the one you are living inside right now.

Footnotes

1. Independent Directors of the Board of Wells Fargo & Company, Sales Practices Investigation Report (April 10, 2017). The report, prepared with Shearman & Sterling, remains one of the most detailed public autopsies of an incentive-system failure ever produced. ↩︎
2. The initial September 2016 consent orders with the CFPB, OCC, and the Los Angeles City Attorney cited approximately 2.1 million accounts; Wells Fargo’s expanded third-party review, announced in August 2017, raised the estimate to approximately 3.5 million. ↩︎
3. Sales Practices Investigation Report, 4 to 8 and 27 to 42. The report describes senior leadership treating widespread integrity terminations as evidence the control system was working, rather than as evidence the incentive system was generating the behavior. ↩︎
4. U.S. House Committee on Transportation and Infrastructure, Final Committee Report: The Design, Development, and Certification of the Boeing 737 MAX (September 2020). The report followed an eighteen-month investigation drawing on some 600,000 pages of records and interviews with Boeing and FAA personnel. ↩︎
5. Final Committee Report, sections on production pressure and what the committee termed a culture of concealment. Internal employee messages released in January 2020 record engineers and pilots expressing alarm about the program’s compromises well before the accidents. ↩︎
6. Anton R. Valukas, Report to Board of Directors of General Motors Company Regarding Ignition Switch Recalls (May 29, 2014). ↩︎
7. Danny Miller, The Icarus Paradox: How Exceptional Companies Bring About Their Own Downfall (New York: HarperBusiness, 1990). ↩︎