Why change initiatives fail (and what makes one stick)

The kickoff Monday

A Monday morning, all-hands meeting. The Transformation is announced: a new system, a new way of working, "as of today we are a different company." There are slides, a new word on the wall, coffee, lukewarm applause. The energy lasts as long as the coffee does.

Six months later, the new system is half-used, the same old spreadsheet survived under the table, the meetings are the same, and no one dares to say out loud what everyone knows: we are back where we were. There was no failure with a bang, no culprit, no date of death. There was a slow and silent restoration of the status quo, as if the organization had antibodies that recognize what is new and neutralize it without a fever.

This article is about why that happens so often, why the usual explanation —"people resist change"— is almost always false, what the evidence says starting by dismantling the most cited and least true statistic in management, and what sets apart the changes that stick from the ones that evaporate.

The 70% myth

"70% of change efforts fail." The line is cited everywhere, almost always attributed to John Kotter and his celebrated 1995 article in Harvard Business Review. There is an awkward problem: in that article the figure is not there. Kotter wrote that he had watched more than a hundred companies trying to transform themselves, and that "a few have been very successful, a few have been utter failures, and most fall somewhere in between, with a distinct tilt toward the lower end." No 70%.

The number appears much later —Kotter only mentions it around 2008— and its most likely origin is another book: the one by Michael Hammer and James Champy on reengineering, from 1993, which estimated "unscientifically" that between 50% and 70% of companies undertaking a reengineering fail to achieve the intended results. From an admittedly unscientific estimate, about reengineering, a universal law about change was born. When the researcher Mark Hughes went looking, in 2011, for the evidence behind the 70%, he found no empirical basis to support it. It is folklore that cites itself.

It would be wise, then, to throw out the number — but not the idea underneath, because that one does have backing, and better backing. McKinsey's global surveys, of thousands of executives, find that fewer than a third of transformations manage both to improve performance and sustain it; in the digital version, barely 16%. But the same study brings the part that really matters: when the company acts on all five fronts of change at once —instead of pulling a single lever— the success rate climbs to 72%.

The conclusion is not "70% fail." It is more useful and less fatalistic: change done by default almost always underperforms; done with method, most of the time it sticks. Failure is neither random nor inevitable. It is a pattern — and patterns can be interrupted.

It is not resistance: it is the design

The comfortable explanation for failed change is psychological: "people resist." It is comfortable because it puts the blame outside whoever designed the change. And it is, almost always, false. People do not resist change —they change phones, cars, houses, even countries—; they resist loss, ambiguity and being changed without a voice or a warning. When a change fails, the cause is rarely in the people. It is in the design. Five typical ways to die:

1. A solution without a problem. The change was born from a book, a conference or a competitor — not from a pain the organization feels as its own. Without a credible "burning platform," the status quo always wins, because it is comfortable and familiar. The first job of any change is to make the option of not changing risky.

2. Confusing the announcement with the implementation. The kickoff is not the change: it is kilometer zero. Between the announcement and reality there are two hundred small operational decisions someone has to sequence, and that no slide contains. Taking the change as done the day it is announced is the most common way of never doing it.

3. No one with skin in the game. The change is handed to a committee that is everyone and is no one, or to a sponsor who gets excited at the launch and then moves on to something else. Without an owner with real authority and something personal to lose if it fails, the change is left orphaned — and orphans, in organizations, do not survive the winter.

4. The incentives keep rewarding the old. A new behavior is asked for and the old one keeps being paid, promoted and applauded. Like the salesperson paid on revenue while being asked to protect the margin, people do what their wallet and their boss reward, not what the poster says. Culture is not decreed: it is the residue of what an organization rewards and tolerates, month after month.

5. Everything at once. The big bang —changing everything, everywhere, on the same day— is the surest way to fail on a grand scale. The data from systems projects is eloquent: large projects succeed in fewer than 10% of cases; small ones, far more. It is no coincidence. What is large cannot be piloted, or corrected, or learned along the way: everything is bet on it going right the first time, and it almost never goes right the first time.

To those five is added a sixth, so silent that it almost deserves its own article —and it has one, the next—: the dashboard did not change. If after the change the company keeps measuring and showing the same numbers as before, it will keep running on the ones from before. What does not start being measured differently does not truly change.

The J-curve: why change gets worse before it gets better

There is a deeper reason why changes are abandoned, and it is almost physical, not psychological. Every real change —a new system, a new process, a new structure— makes performance worse before making it better. There is a period in which people can no longer use the old and do not yet master the new: they stumble, they get frustrated, they produce less. It is the J-curve, or "the valley of despair."

The problem is that the bottom of the J is exactly where "this does not work, let's go back to how we were" is heard loudest. And going back, in the pit, is tempting: it eases the pain immediately. The organizations that give up there never see the climb, and draw the worst possible lesson: that change is useless. Next time they start with less credit, and the cycle repeats until "changes don't work here" becomes a self-fulfilling prophecy.

The dip is not the sign that the change failed; it is the sign that it began.

How to make it stick (without theater or bureaucracy)

Change that works is not louder: it is more designed. The acid test of every gesture is the same one that holds for all professionalization: does this bring the change closer to the real operation, or does it only decorate it? Five moves, in order.

1. Start with the problem, not the solution. Name the pain, measure it and make it visible before proposing anything. A change that people understand as a response to a real problem that affects them does not need to be "sold": it explains itself. The energy spent on convincing is usually the measure of how badly the problem was chosen.

2. An owner and a small coalition with power. Not a committee: a person in charge with authority and something to lose, plus a small group that pushes from different areas. Kotter called it the guiding coalition; the name matters less than the idea — change needs concentrated political muscle, not consensus diluted among twenty people who nod and none who answers.

3. Pilot rather than go big bang. One area, one process, one piece. Learn, correct, and only then extend. The pit of the J is survivable in a pilot and lethal across the whole company at once. And there is an extra benefit: a pilot that works is the best argument there is. It convinces more than any presentation, because it already happened.

4. Move the incentives and the routines, not the posters. Change what is measured, what is paid and who gets promoted; change the cadence of meetings and who decides what. Behavior follows incentives and routines; culture comes behind, never ahead. Motivational posters are the cheap —and empty— substitute for that work.

5. Change the dashboard. If you want a new behavior, you have to measure the new and stop celebrating the old. It is the bridge between the change and day-to-day management: without new numbers, the change has nothing to hold on to when the enthusiasm fades.

And above the five, two disciplines that decide the rest. Communicate the why many more times than seems necessary — Kotter observed that almost all changes are undercommunicated by a factor of ten; what for leadership is already obvious and repeated, for everyone else is only beginning to be heard. And protect the change through the pit of the J, knowing in advance that it will hurt, warning about it, and resisting the temptation to reverse course just when it hurts most — which is also when it is about to improve.

The difficult conversation

"People resist change." People resist loss and ambiguity, not change. Give certainties, explain the why, give a voice, and take care of whoever loses something — and the resistance drops on its own. Almost always, resistance is information about a badly designed change, not about difficult people. It is worth listening to before defeating it.

"We need a cultural change." Culture is not decreed nor transformed in a two-day workshop. It is what remains after months of rewarding certain things and tolerating others. Change the incentives and the routines, sustain it, and culture changes behind. The other way around never works: no poster beats a commission system.

"We already announced it." Announcing is not changing. The announcement is kilometer zero of a route that almost no one travels to the end. The boardroom question is not what was announced, but what was sequenced, who sustains it, and what will happen when we reach the pit.

Changes do not die of resistance. They die of orphanhood, of haste and of dashboards that did not change. None of those three causes is about the people — all three are decisions. And decisions can be made better.