Change Management: Why Change Fails in the Middle, Not at the Start

The day after the town hall

Picture the launch. The CEO stands in front of the whole company, the new operating model glowing on the screen behind her. People clap. The slides go out by email that afternoon. Six months later, a finance analyst is still rebuilding the forecast in the spreadsheet the new system was meant to retire, because the spreadsheet is faster and her manager still asks for it. Nobody decided to resist. The old way simply kept working better for the person doing the work.

That is where most change goes wrong. Not at the announcement, which is usually well rehearsed, but in the long middle, when the new way is still clumsy, the old way is still available and the people who were asked to change have every practical reason not to.

Change management is the work of moving people, not just plans, from an old way of working to a new one: making the case, equipping people to work differently, removing what holds the old way in place and checking, long after the launch, that the new behavior has actually stuck. It covers the human side of a reorganization, a new system, a merger, a strategy or a new process.

Announcements start changes. Middles decide them.

This page covers what change management is, the models people use and what each is good for, a statistic you should stop quoting, and a plan built for the months after the launch rather than the day of it.

Two disciplines with the same name

The phrase means two different things, and it helps to know both. In organizations it means the people side of change: getting a few hundred or a few thousand people to work differently. In plant and process safety it means a formal control. The US Occupational Safety and Health Administration requires employers covered by its process safety standard to "establish and implement written procedures to manage changes" to process chemicals, technology, equipment and procedures, and to address, before any change, its technical basis, its impact on safety and health and who must authorize it ⁸.

That rule has a history. On June 1, 1974, the Nypro plant at Flixborough in England was destroyed by an explosion that killed 28 workers. A leaking reactor had been removed and a bypass pipe installed so that the plant could keep producing. The UK Health and Safety Executive's summary is blunt: "A plant modification occurred without a full assessment of the potential consequences," no calculations were done for the new line, and "No drawing of the proposed modification was produced" ⁹.

The two disciplines share one insight. The danger is rarely the change everyone is watching. It is the change made in a hurry, in the middle, by people trying to keep things running, that nobody stopped to examine. Process safety writes that into procedure. Organizational change management has to build the same habit into how leaders run the months after a launch.

The models, and what each is for

There is no shortage of change models. They answer different questions, and choosing one is mostly a matter of knowing which question you are asking.

Two of those deserve a footnote that most training decks skip. The three-step model credited to Kurt Lewin is largely a later construction. Stephen Cummings, Todd Bridgman and Kenneth Brown traced it back and concluded that Lewin "never developed such a model and it took form after his death"; in his 1947 article the idea is "buried there in the 24th of 25 sub-sections in a 37 page article" ⁵. And the "change curve" of denial, anger and acceptance comes from Elisabeth Kübler-Ross's work with dying patients. Her foundation notes that she later extended it to other losses, workplaces included, but that she "did not design a management model" ¹¹. Both ideas can still be useful. Neither carries the authority its popularity implies.

For the detail of each model, see the ADKAR model and Lewin's change model. For the communication plan that sits alongside any of them, see the change communication strategy.

Why the middle is where change dies

John Kotter's 1995 Harvard Business Review article is still the best short account of how transformations fail, and its most important sentence is near the end: "Until changes sink deeply into a company's culture, a process that can take five to ten years, new approaches are fragile and subject to regression" ¹. Two of his eight errors live in the middle: not planning for short-term wins, and declaring victory too soon. In one case he describes, "Within two more years, the useful changes that had been introduced slowly disappeared" ¹.

The middle is hard for a structural reason. At the launch, the costs of the change are visible and its benefits are promises. For a while, the new process is slower, the new system is unfamiliar and the old way is one click away. Every individual decision to fall back is reasonable. Added together, they are the change failing. William Bridges drew the line that explains it: change is the external event, while transition is "the inner psychological process that people go through," and it starts with an ending, then passes through a neutral zone, before anything new begins ¹⁰.

The other reason is that bad news travels slowly in the middle. At Ford in 2006, executives presented status charts that were all green while the company was heading for a $14.6 billion net loss in 2008 ¹⁶. "We had a culture where, if you reported bad news, you were done," Mark Fields later said. When he showed a chart with red on it, flagging problems with the coming Edge crossover, "All the chairs moved away from me." Then the new chief executive, Alan Mulally, started clapping. Fields called that meeting "a tipping point at Ford" ¹². A change that cannot hear about its own problems cannot fix them in time.

The 70% figure, traced

Almost every article on change management opens by saying that 70% of change programs fail. The figure usually traces to Michael Beer and Nitin Nohria's 2000 Harvard Business Review article: "The brutal fact is that about 70% of all change initiatives fail" ⁶. It was an assertion, not the result of a study.

Mark Hughes went looking for the evidence. Reviewing five published instances of the claim, he concluded that "there is no valid and reliable empirical evidence to support such a narrative" ⁷. Kotter's own 1995 article, often cited as the source, does not give a number at all; he wrote that a few efforts succeeded, a few failed utterly, and "Most fall somewhere in between, with a distinct tilt toward the lower end of the scale" ¹.

That last sentence is the useful one. Most change does not fail cleanly. It half-succeeds: the new system goes live and the workarounds go on, the reorganization happens and the old decision habits survive it. The same pattern applies to strategy itself; the 70% strategy failure figure has an equally thin history.

A change plan for the months after launch

A launch plan is the part most organizations already have. These steps cover what usually goes missing between the launch and the point where the new way is simply how work is done.

1. Name what is endingWrite down what people lose: a tool they are fast in, a team, a status, a way of being judged. Bridges' point is that people let go of the old before they take hold of the new, so say plainly what is ending and why ¹⁰.
2. Measure behavior, not activityTraining completed and go-live reached are activity. Share of work done the new way, and how many people still use the old tool, are behavior. Track the second set weekly through the middle months.
3. Diagnose each stuck groupFor any team that has not moved, ask the ADKAR question: is the gap awareness, desire, knowledge, ability or reinforcement? Each needs a different fix, and a training course only fixes one of them ³.
4. Make the old way harderKotter's fifth error is not removing obstacles to the new vision ¹. Retire the old report, close the old channel, stop asking for the spreadsheet. If leaders still request the old output, the change has not been made.
5. Plan the early wins on purposePick results that can show up within months, assign them owners and announce them when they land. They buy patience for the slower parts of the change ¹.
6. Keep a channel for redAsk every review for one thing that is going wrong, and thank the person who brings it. A status report that is always green is a warning, not a comfort ¹².
7. Hold the victory lapDo not declare success at go-live. Declare it when the behavior numbers have held for two consecutive review cycles, and keep the measures running after that.

A worked example

The company below is invented, to show the steps working together.

Notice that nothing in the fix is a communication campaign. The handlers already understood the change. What kept the old way alive was a slower screen and a meeting that still ran on the spreadsheet.

Microsoft: changing how people argue

Culture change is the hardest case, because the old way is not a tool you can switch off. When Satya Nadella became Microsoft's chief executive in February 2014, his first email to employees said that "Our industry does not respect tradition — it only respects innovation," and asked that every employee "help drive cultural change" ¹³.

He was specific about what had to end. Describing the old culture of Microsoft's meetings, he said that "The idea was to destroy anyone's [argument] in the first five minutes by asking highly precise questions," a habit that was "being used as an instrument of offence" ¹⁴. In its place he described a "culture of growth mindset" in the company's 2017 annual report ¹⁵. Our analysis of the Nadella turnaround looks at how much of what followed came from culture and how much from the cloud business he inherited.

The lesson for anyone running a change is the first move, not the result. Nadella named the behavior that was ending, in the words people used for it. A change that only describes the destination leaves everyone to guess what they are supposed to stop doing.

Before you launch: a checklist

• We have written down what is ending for each affected group, not just what is beginning.
• Every affected group has a named leader who will be measured on adoption, not on go-live.
• We have two or three behavior measures that show whether work is being done the new way, with a baseline.
• We know which old tools, reports and meetings will be retired, and on what dates.
• Training happens before go-live, close enough to it to be remembered.
• Each review asks what is still being done the old way, and why.
• There is at least one early win, with an owner and a date, inside the first three months.
• We have agreed in advance what result, held for how long, will count as success.

Monday morning: pick the change your organization launched most recently and ask the people closest to the work what they are still doing the old way. The answer is your plan for the next quarter.

Change rarely fails loudly. It fades, one sensible workaround at a time, while the status report stays green. The work of change management is to stay in the room after the applause, long enough to see which way the middle is going.

Questions

What is change management?

Change management is the work of moving people from an old way of working to a new one: making the case for the change, equipping people to work differently, removing what holds the old way in place, and checking long after the launch that the new behavior has stuck.

What are the main change management models?

The most used are Kotter's eight steps for organization-wide change, Prosci's ADKAR model for diagnosing where an individual is stuck, and William Bridges' transition model for the psychological side of change. Lewin's "unfreeze, change, refreeze" is widely cited, but researchers have shown Lewin never presented it as a model.

What is management of change (MOC) in process safety?

In process safety, management of change is a formal control. OSHA's process safety standard requires covered employers to have written procedures for changes to process chemicals, technology, equipment and procedures, addressing the technical basis, the safety impact and who authorizes the change before it is made.

Do 70% of change initiatives fail?

The figure is widely repeated but not well founded. It usually traces to a 2000 Harvard Business Review article, and a 2011 review by Mark Hughes found no valid and reliable empirical evidence for it. Most change half-succeeds rather than failing outright.

What is the difference between change and transition?

In William Bridges' terms, change is the external event, such as a new system or structure, while transition is the inner psychological process people go through to accept it: an ending, a neutral zone and a new beginning.

Where to go next

• Strategy Execution: Where It Breaks Between the Offsite and Monday
• Why Strategies Fail, and Where the “70%” Figure Comes From
• ADKAR Model
• Change Fatigue
• Change Communication Strategy