The Durable Organization

THE BOTTOM LINE

Strength and durability are different things. A system can look strong because it refuses to change, and that refusal is what makes it brittle.

Political history shows the pattern. Authoritarian states are not weak, and some last a very long time. But they often turn change into a break. In autocracies since the Second World War, about 45 percent of leadership changes brought regime change with them. By Stephen Kotkin’s account, the Soviet Union was stable into the mid-1980s, and the attempt to reform it is what ended it. Established democracies look noisier, and they are among the most stable systems as well. They have channels that turn pressure into small corrections.

Organizations show the same pattern, and here it can be measured. Change carries a risk of its own. In one long study, each change raised the odds of failure for a time. Yet firms that change at a steady rhythm have outperformed firms that change in bursts, and declining firms that act early have recovered more often. Change that is put off does not go away. It arrives later as one large change, under pressure.

Change is often put off for two reasons. The signals that call for it are not on the instruments leaders watch. And many leaders believe that people resist change, a belief the evidence supports far less than its reputation suggests.

This paper does not argue that companies should be run like democracies. It argues that a durable organization keeps those signals in view and changes at the speed of need.

PART ONEStrong is not durable

Most people judge the strength of a system by how much it controls. A state looks strong when its leader has no rivals, its streets are quiet, and its orders are carried out. A company looks strong when it leads its market, its revenue is high, and no competitor is close. In both cases, strong is taken to mean safe.

The record does not support that. It does not support the opposite either.

Start with states. It is tempting to say that dictatorships are fragile and democracies are stable. The data says something else. A forecasting study led by Jack Goldstone, published in 2010, examined political instability in countries around the world from 1955 to 2003. The risk of instability was lowest at the two ends, in full autocracies and full democracies. The least stable were partial democracies split into factions. Their odds of instability were more than thirty times those of a full autocracy.

So authoritarian states are not weak, and some last a very long time. Soviet rule lasted seventy-four years.

What differs is the form that change takes. In a democracy, a change of leader is routine. The government changes and the system stays. Autocracies are different. Barbara Geddes, Joseph Wright, and Erica Frantz assembled a record of the 280 autocratic regimes that existed between 1946 and 2010. About 45 percent of the leadership changes in those regimes led to a change of regime. The rest did not, and the same record shows that some kinds of autocracy hand power on more smoothly than others. Still, in such systems, replacing the person at the top often puts the whole structure at risk.

Stephen Kotkin, the historian of Russia, made the point in a 2022 interview with The New Yorker. Authoritarian rule, he said, is “all-powerful and brittle at the same time.”

That line holds the idea of this paper. The features that make such a system strong are the same features that can make it brittle. A ruler who removes every rival can leave no orderly way to replace him. A system that punishes bad news tells its leaders less and less of the truth. Kotkin makes this point as well. A system that holds everything in place leaves change nowhere to go until it comes all at once.

Engineers use two words for what most people treat as one. Strength is the load a material can carry before it gives way. Toughness is how much it can absorb before it breaks. Glass resists great pressure and has very little toughness. It gives no warning before it shatters. Steel bends first. The bend is a warning, and the steel is still in one piece.

This paper uses the word durable for that second quality. A durable system can take change and keep going.

Established democracies are durable in this sense, and the reason is not that they are calm. They are noisy. This paper’s reading is that the noise comes from working channels. Elections, courts, a free press, and open argument carry pressure to the people in charge while the pressure is still small. They force small corrections. The name democracy does not do this work. The channels do. Where the channels clog, a democracy stores up change like any other system. An earlier commentary by this paper’s author looked at one form of that clogging, in the administrative machinery of democratic governments. Its title is We Gave Bureaucracy the Fastest Tools in History. It Got Slower.

Organizations are judged the way states are. Market share, revenue, headcount, and market value measure strength at one moment. They do not show whether the organization can change. Three companies show the gap.

Digital Equipment Corporation was the second largest computer maker, behind only IBM. Its sales reached 14 billion dollars. In 1998 Compaq bought what was left of it.

Nokia led the world market for mobile phones for more than a decade. In the second quarter of 2008 it put its own share of that market at 40 percent. In April 2014 it completed the sale of its phone business to Microsoft.

General Motors held 54 percent of the American car market in 1954. In June 2009 it filed for bankruptcy.

None of these companies was weak at its peak. Two led their fields, and the third was second only to IBM.

Digital shows the pattern most plainly. Edgar Schein of MIT consulted to the company from 1966 to 1992 and later wrote its history with three co-authors. Digital’s culture prized creativity and invention, and that culture built the company. By Schein’s account, the same culture later became so fixed that Digital could not adapt, even though it saw the need. What made it strong made it brittle.

Nokia and General Motors return later in this paper. Each shows a different part of the argument.

This paper does not claim that authoritarian states are weak, or that any of them is close to its end. It does not claim that a democracy is safe because of its name. The claim in this part is narrower. Strength and durability are separate qualities, and the usual measures capture only the first.

The next part takes up what happens when change comes late.

PART TWOLate change is the dangerous change

If strong systems are brittle, the obvious advice is to change. That advice is too simple, because change is dangerous too. This part holds both facts together.

Start again with states. In Armageddon Averted, published in 2001, Stephen Kotkin argues that the Soviet Union did not have to end when it did. As late as 1985, in his words, it was “lethargically stable.” It had lost the competition with the West, and it could still have carried on for a long time. What ended it, by his account, was the attempt to reform a system that could not be reformed. Neither the leaders nor the public expected reform to end in collapse.

Alexis de Tocqueville saw the same pattern in France more than a century before. Writing in 1856 about the fall of the old monarchy, he concluded that a bad government is usually in the greatest danger at the moment it starts to reform. His explanation was about people. They put up with a burden for as long as it seems unavoidable. Once relief looks possible, the same burden becomes unbearable.

Neither writer argues against reform. This paper reads the two cases for what they share. Neither system had stood entirely still. But each had gone a very long time without the change it most needed. When that change was finally attempted, it was very large, and the system had no practice at change of that size.

A caution belongs here. A state is not a company, and Tocqueville’s explanation is about how a population reacts. The link this paper draws between states and organizations is an analogy. For organizations there is direct evidence, and it points the same way.

Terry Amburgey, Dawn Kelly, and William Barnett studied 1,011 Finnish newspapers over 193 years. They found that a change in a newspaper had two effects. It raised the risk that the paper would fail, and it did so right away. It also made another change of the same kind more likely. Both effects faded with time.

Put those findings together. Every change carries a risk. The risk passes if the organization survives it. And an organization that has just changed is more likely to change again.

From here this paper proposes a mechanism. A small change carries a small risk, and each one keeps the organization in practice. A change that is put off does not go away. The need grows. When the change finally comes it is large, and a large change carries a large risk. It also comes at the worst time. The organization is out of practice. Money is short. And the team at the top is often weaker than it was. The studies cited here do not test this mechanism directly. In particular, none shows that practice at small changes makes later change safer.

Earlier papers in this series describe two parts of this mechanism. Nothing Stays Tuned explains why the need grows even when nobody decides anything. An organization’s commitments drift, and the gap widens unless someone is watching it. Fast to Cut, Slow to Build describes how large companies fall out of practice. They stay quick at cutting, and they slowly lose the ability to move money and people toward what they need next.

The point about the team at the top has evidence behind it. Donald Hambrick and Richard D’Aveni compared 57 large companies that went bankrupt with 57 similar companies that survived. The top teams of the failing companies differed from those of the survivors, and the gap grew wider and faster over the last five years before bankruptcy. The authors propose a spiral that runs both ways. A flawed team makes the company’s trouble worse. The trouble then weakens the team. Executives leave, others are blamed and removed, and the company can no longer attract strong replacements.

General Motors shows what late change looks like. Its bankruptcy came at the end of a decline that had lasted three decades. The decline was not hidden, and GM had not stood still. It kept bringing out new products and cutting costs. None of it stopped the slide. GM’s share of the American market fell from 27.3 percent in 2004 to 19.5 percent in mid-2009. By April 2009 it had lost 82 billion dollars since 2004.

Then the largest change came all at once. GM filed for bankruptcy on June 1, 2009. A new company began operating on July 10, less than six weeks later. It kept four core brands in the United States and moved to shed four others: Pontiac, Saturn, Hummer, and Saab. More than 40 billion dollars of obligations were removed. The cuts did not end in court. GM said its American workforce would fall from about 91,000 at the end of 2008 to about 64,000 by the end of 2009, with plant and dealer closures to follow. The rescue took about 60 billion dollars of public money, and the governments of the United States, Canada, and Ontario owned 72.5 percent of the new company.

A restructuring forced by decades of decline took less than six weeks in court. GM did not control it. GM survived because governments paid. Most organizations have no one to do that.

The general idea here is not new, and credit is due. Nassim Taleb and Mark Blyth argued in Foreign Affairs in 2011 that a system whose normal ups and downs are artificially held down becomes very fragile while showing no visible risk. Taleb’s image is a forest. Put out every small fire, and the first large fire that cannot be controlled destroys the whole forest. Gary Hamel and Liisa Välikangas made the point about companies in Harvard Business Review in 2003. A turnaround, in Hamel’s words, is “transformation tragically delayed.”

What this paper adds is the join between the two fields, the evidence on timing in the next part, and an explanation of why change gets put off in the first place.

So the choice is not between change and safety. It is between small risks taken on time and one large risk taken late. Hold, Collide, Come Apart, another paper in this series, made the same point about the divisions inside a company. The same correction is cheap early and expensive late.

The next part looks at what steady change buys, and where that evidence has limits.

PART THREEWhat steady change buys

Part Two argued that small risks taken on time are better than one large risk taken late. That is an argument. This part asks what the evidence says. Some good evidence supports the argument, and some evidence complicates it. Both belong here.

The strongest support comes from a study of rhythm. Patricia Klarner and Sebastian Raisch followed 67 European insurance companies from 1995 to 2004. They tracked each company’s major changes in strategy and looked at the pattern over time. Some companies changed at regular intervals, with periods of change and periods of stability taking turns. Others changed irregularly. The companies that changed regularly outperformed the ones that changed irregularly. The result held under different conditions inside and outside the companies, and across time periods.

Two details of that study matter. First, the finding is about regularity. It is not about changing as often as possible. The authors treat how regularly a company changes and how often it changes as separate things with separate effects. Second, the successful pattern included stability. Their conclusion is that a regular balance between change and stability goes with long-term success.

The second piece of support concerns companies already in decline. Chanchai Tangpong, Michael Abebe, and Zonghui Li studied 96 American firms in matched pairs. Declining firms that began to cut back early were more likely to turn around.

Now the complications.

The first is inside that same study. What a firm did early mattered. Selling off businesses early helped. Leaving geographic markets early helped. Laying people off early did not. So “act early” is too simple. This paper’s reading is that the early moves that helped were changes to the business itself.

The second comes from a larger study. José Luis Barbero and two colleagues examined 263 declining American firms over 26 years, from 1983 to 2009. Early cutbacks helped when the firm’s market was generous, with resources to spare. Early cutbacks hurt when the market was changing fast. In those fast- changing markets, what helped was speed once the cutting began. And in this study an irregular rhythm of cutbacks did better than a regular one. The authors’ overall conclusion still favored acting with urgency. But the details do not line up neatly with the insurance study. The two studies also measured different things. One looked at changes in strategy across a whole industry. The other looked only at cutbacks in firms already in decline.

The third is on the side of states. Nassim Taleb and Gregory Treverton argued in Foreign Affairs in 2015 that the best sign of a country’s future is recent moderate volatility, and that a long calm is a poor one. That is this paper’s idea in its boldest form. A reply by Paul Stares in the same journal said the authors had not supported the claim, and that their method did no better than the warning methods already in use. This paper does not rest on that claim.

The fourth is a limit on all of these studies. Each covers one kind of firm in one period. Each shows that two things went together. None shows that one caused the other. Companies that change at a regular pace may simply be better run.

One finding from Part Two also stands over everything here. In the newspaper study, each change raised the risk of failure. Change is never free. An organization that changes for the sake of changing takes on risk and gets nothing for it.

So what does the evidence support? Three modest statements. First, a regular pace of change, with stability in between, went with better long-term performance in the study of rhythm. Second, for firms in decline, acting early on the business itself went with recovery, at least where the market was not changing fast. Third, nothing here supports constant change, and nothing here supports early layoffs.

The evidence does not say to change more, or to change faster. It says not to let the need for change run far ahead of the change itself. This paper calls that changing at the speed of need. The phrase comes from an earlier commentary by this paper’s author, which set the speed of need against the speed of red tape. The need sets the pace. Slower than the need, and change piles up. Faster than the need, and the organization pays the risk of change for no reason.

That raises the question the rest of this paper is about. If changing on time is better, why do capable leaders so often change late? The usual answer is that they refuse to. This paper gives two other answers. The next part argues that more often they do not see the need in time. The part after it takes up a belief about people that makes delay look wise.

PART FOURWhy change gets put off

There are two common explanations for late change. One is that leaders see the need and refuse to act. The other is that they do not see the need in time.

Research on decline has room for both. A model published in 1989 by William Weitzel and Ellen Jonsson describes decline in five stages. In the first, the organization fails to notice the early signs of trouble. In the second, its managers recognize the need for change and still do nothing. Refusal is the second stage. Not seeing comes first.

This paper’s view is that not seeing is the more common of the two. This paper has not found a study that counts how often each occurs, so this is a judgment and is offered as one.

The reason lies in what leaders watch. Revenue, market share, costs, and output are the standard instruments. All of them report results, and results arrive last. Nokia put its share of the world phone market at 40 percent in the second quarter of 2008. The iPhone had gone on sale the year before.

An earlier paper in this series, The Unmeasured Layer, argued that a large organization sees its own working condition only in fragments. Different functions each hold a piece. The pieces are rarely assembled into one standing view. How decisions really get made and how work really moves are known inside the organization. They are seldom brought together on the page that leaders read.

Trenches, Not Silos, another paper in the series, made a related point. Companies watch customer counts, revenue, and product offerings because those numbers are easy to count. The problems that are hard to count are left to grow.

Nokia shows how this plays out, and here the evidence is unusually good. Timo Vuori and Quy Huy conducted 76 interviews with Nokia’s top managers, middle managers, engineers, and outside experts about the years 2005 to 2010. Top managers saw the outside threat clearly. They were afraid of competitors and of shareholders. Middle managers were afraid of something closer, which was their own superiors and peers. Out of that fear they stayed quiet or sent up optimistic reports. Top managers came to believe that Nokia’s technology was stronger than it was, and they pushed harder on that basis. The authors do not say fear was the only cause. They say it was an important one.

Nokia’s leaders did not ignore the threat. They saw the competitor. What they could not see was their own company.

“They saw the competitor. What they could not see was their own company.”

Part One described the same pattern in a state. A ruler who punishes bad news hears less and less of the truth. The mechanism does not need a dictator. It needs only that telling the truth to the people above carries a cost.

That cost is common. Frances Milliken, Elizabeth Morrison, and Patricia Hewlin interviewed 40 employees in a range of industries. Of those, 85 percent said they had at some point held back a concern they thought was important and had not raised it with a supervisor. The study is small and exploratory. Its point is still plain. The knowledge that something is wrong often exists inside an organization and does not travel upward.

There is a second way an organization can learn it is in trouble, and it is quieter. Albert Hirschman described two ways that management finds out about its own failings. People speak up, or people leave. He also gave a warning. The people who care most about quality are the ones best placed to speak up, and they are also the most likely to leave. This paper adds one observation. In a company, the people who can leave most easily are the ones most in demand. When they go, the company loses their work and their warning at once. Their departure is a signal. It is usually recorded as turnover and handed to recruiting.

Exit raises a larger argument that this paper does not take up. The philosopher Elizabeth Anderson has compared the modern workplace to a private government, and she holds that the freedom to quit does not change that. Exit may count for more than she allows. People do change jobs, and companies differ a great deal in how they are run. But she is right that exit is uneven. It is most open to those with the most options. For this paper the point is narrower. Each such exit is information, and most organizations do not read it.

So the signs that change is needed do exist. They are in what employees know and do not say, in who leaves, and in how the work really moves. These are the vital signs of an organization’s durability. Its results are the signs of its strength. Leaders mostly watch the second kind.

This is how not seeing produces late change. By the time the results show the problem, the need is already old, and the change required is already large. Nothing Stays Tuned put the principle in a few words: seeing comes before doing.

There is a second reason change gets put off. It is a belief about people, and the next part takes it up.

PART FIVEWhat people actually resist

There is a second reason change gets put off. Many leaders believe that people resist change, and that any change will be a fight. A leader who believes this has a reason to make changes rarely, and only when forced. The belief deserves a close look, because the evidence for it is weaker than its reputation.

Start with where the idea came from. Eric Dent and Susan Galloway Goldberg traced its history in 1999. The term goes back to Kurt Lewin, and he did not mean it as a fact about employees. He described resistance as a force in the whole system, one that acts on managers and employees alike. Later writers kept the phrase and dropped the meaning. Resistance became a flaw in employees that managers had to overcome. Dent and Goldberg’s conclusion is that people do not resist change itself. They resist losing status, pay, or comfort.

The first major study of the subject points the same way. In 1948 Lester Coch and John French studied a pajama factory where production workers were resisting changes to their methods and jobs. The researchers introduced a change to four groups in different ways. One group had no part in planning it. The others took part. The sharpest difference in results was between the group with no part and the groups that took full part. A later review of that work stresses its real claim, which is that resistance comes from the setting in which a change is made and not from the individual.

More recent work points the same way, with a caveat. In four studies, Alexandra Michel, Rune Todnem By, and Bernard Burnes examined how a personal tendency to resist change, and the features of a change itself, related to commitment to that change. People were more committed to changes they saw as beneficial. With one exception, the personal tendency did not alter that link. The authors conclude that the level of resistance to any one change depends on the setting and on how the change is managed. They also note that a person’s disposition shapes how change is viewed.

Then there is the famous number. It is often said that 70 percent of change efforts fail. Mark Hughes reviewed five published sources for that figure in 2011 and found no valid and reliable evidence behind it. The figure appears to have begun as an estimate in a 1993 book about one kind of change, and the authors of that book called their own estimate unscientific.

None of this means that people welcome every change. Experiments by William Samuelson and Richard Zeckhauser showed that people stay with what they already have more often than the choices in front of them justify. A later replication confirmed the effect in three of four tests. The lean toward the familiar is real. It is a lean, though. The same body of research shows people accepting change when they see a benefit in it and have a part in it.

There is also direct evidence that people differ. In 2003 Shaul Oreg built and tested a scale that measures a personal inclination to resist change. Scores on it predicted how people reacted to particular changes. One of its four parts is the reaction to change that is imposed.

So the common statement is too broad. Some people are more averse to change than others, and that tendency can be measured. But resistance is not a fixed fact about employees. What people resist most reliably is loss, and change that is done to them.

What, then, is hard on people? Alannah Rafferty and Mark Griffin identified three features of change that shape how employees react: how often it comes, how large it is, and how well it was planned. In research along these lines, change that was planned went with less uncertainty. Change that came very often went with more uncertainty and with more people intending to leave.

A survey record tells a similar story. Gartner reports that in 2016 the average employee went through two planned company-wide changes, and 74 percent of employees were willing to support such change. In 2022 the average employee went through ten, and the willing share had fallen to 43 percent. Two things stand out. About three in four people were willing when the load was light. And willingness fell as large programs piled up. The pandemic years fall inside that period, so the numbers should be read with care.

Put together, the evidence says people are most sensitive to change that is large, sudden, imposed, or piled on. This paper found no study that measures delay itself. But a change that has been put off for years tends to arrive in exactly this form. It is large because the need has grown. It is sudden because a crisis forces it. It is imposed because there is no longer time to involve anyone.

That suggests a loop, and this paper offers it as a proposal. A leader expects resistance and puts off small changes. The need grows. The change finally arrives large and sudden. People react badly. The leader takes the reaction as proof that people resist change. The belief is confirmed by the delay it caused.

The next part turns to what the late options cost, and to what a durable organization does instead.

PART SIXWhat a durable organization does

The argument so far comes to this. Strength and durability are different qualities. This paper argues that change is safest when it is small and on time, and most dangerous when it is large and late. Change comes late mostly because leaders do not see the need, and partly because they expect more of a fight than the evidence predicts.

The lesson follows. Watch the organization’s vital signs, and do not judge its health by its strength on paper. Make changes at the speed the organization requires. Do not let them build up until the only tools left are the largest ones.

In practice that means four things. They are recommendations, and they go beyond what the studies prove.

First, read the vital signs. Ask the people who do the work what is going wrong, and make it safe to answer. Nokia’s top managers were not short of information. They were short of true information. Treat the departure of able people as a signal, and ask what they saw. Look at how decisions and work really move, and put that picture in front of the people who run the organization. The Unmeasured Layer proposes what that picture should cover.

Second, set a regular rhythm for looking. The need for change is hard to see, so a durable organization does not wait until it is obvious. It reviews its own condition on a schedule, changes what the review finds, and leaves the rest alone. This is how a regular pace and the speed of need fit together. The regular review is what keeps the organization from falling behind the need. Nothing Stays Tuned describes such a practice. It starts with the commitments that matter most and get the weakest feedback. It uses calendar dates as checkpoints, and it adds a review when conditions change or warning signs appear.

Third, make changes while they are small, and give the people affected a part in them. The research in Part Five points to both. People accept change more readily when they see a benefit in it and have a hand in it.

Fourth, treat the largest tools as last resorts, and know what they cost.

Reorganizations are the first of these. In a McKinsey survey of 1,800 executives, about two thirds of reorganizations delivered at least some improvement. But more than 80 percent failed to deliver the value they were meant to in the time planned, and about 10 percent did real damage to the company. The same article reports that productivity fell noticeably in about 60 percent of cases. A Bain study of 57 reorganizations found that fewer than a third produced any meaningful gain in performance.

Large layoffs are the second. They are also the easiest of these tools to reach for. Fast to Cut, Slow to Build noted that a large company can plan a round of layoffs in a few weeks. The costs come later. In a study of 200 companies by Charlie Trevor and Anthony Nyberg, downsizing was followed by more voluntary quitting. In some cases the people who quit afterward far outnumbered the people laid off. The cost is not only to the company. A Finnish study followed 22,430 public employees who kept their jobs through a period of cuts. Where the cuts were largest, sick leave rose among permanent staff, and deaths from cardiovascular disease were about twice as frequent as where there were little or no cuts. And in the study of declining firms described in Part Three, early layoffs did not improve the odds of recovery.

Company-wide transformations are the third. Here this paper gives no number, because the number usually quoted is the 70 percent figure, which has no valid and reliable evidence behind it.

These tools share one thing. Each is large, sudden, and imposed. Each is the form that change takes when it has been put off.

Sometimes a large and urgent change cannot be avoided. A shock comes from outside, or the need was truly hidden, or a new leader inherits years of delay. Then the change should be made, and made without illusions. The larger study of declining firms found that, once a firm was in decline, acting with urgency generally went with better results. But the risks described in this paper still apply. A change raises the chance of failure for a time, and a large one is hard on the people inside. Urgency does not remove those costs. It only makes them worth paying.

This paper has limits, and they should be stated plainly.

The link between states and companies is an analogy. The pattern in states is a tendency and not a law. Some authoritarian systems have made very large changes and lasted, and single-party regimes in particular have proved long-lived. The forecasting study covers 1955 to 2003, and a later test of its model found that its accuracy varied over time. So the regime findings describe a period. They are not a permanent ranking.

The studies of organizations are few. Each covers one kind of organization in one period, and each shows what went together, not what caused what. The three company cases illustrate the argument. They were chosen because of how they ended, so they cannot prove it.

Two claims are this paper’s own judgment. One is that not seeing is more common than refusing. The other is the loop in Part Five, in which the belief in resistance causes the delay that seems to confirm it.

One limit needs its own note. Changing on time does not guarantee that a change lasts. The best long-run evidence comes from a study of Indian weaving firms. Nine years after an experiment that improved their management practices, about half of the practices the plants had adopted had been dropped, though those plants still used more of the practices than the comparison plants in the experiment. The reasons given most often were that managers had moved on and that directors lacked the time. This paper’s author is currently researching what causes change to fade. The studies available have limits. Few follow an organization for many years after a change. In a first review of 78 records, the author found three studies that clearly met the review’s rules. None of the three recorded, before the work began, who inside the organization stood to lose from it. Until that is better understood, the advice in this paper comes with a caution. Change made on time is better than change made late. It still has to be kept.

The claim that remains is the one this paper began with. A system can be strong and still be brittle. What makes an organization durable is not how firmly it holds its position. It is whether it can see its own condition and change while the change is still small. A durable organization keeps its vital signs in view and changes at the speed of need.