Seventy percent of change programmes fail. The figure appears in consulting decks, in the opening slide of change management training, in the introductions of academic papers and in the marketing copy of every firm that sells transformation support. It is precise enough to sound like a measurement and round enough to be memorable, and it has been in circulation for more than thirty years.
In 2011 an organisational change researcher at the University of Brighton went looking for the study behind it. There is no study behind it.
What the review actually did
Mark Hughes published the search in the Journal of Change Management under a title that leaves little room for doubt about the finding: Do 70 Per Cent of All Organizational Change Initiatives Really Fail? The method was straightforward. He took five separate published instances of the claim, followed each one back to whatever it cited, and assessed the evidence at the end of the chain.
Whilst the existence of a popular narrative of 70 percent organizational change failure is acknowledged, there is no valid and reliable empirical evidence to support such a narrative.
In each of the five cases the trail ended the same way. Either the source stated the figure without supporting evidence, or it cited another source which in turn stated the figure without supporting evidence. At no point in any of the five chains was there a dataset, a sample, a definition of failure or a method.
The 1993 origin, and what its authors said about it
The earliest version most commonly pointed to is in Michael Hammer and James Champy's Reengineering the Corporation, published in 1993. Their claim was that somewhere between 50 and 70 percent of reengineering efforts did not achieve the results intended, and they characterised it as an unscientific estimate.
That is a careful and honest thing for an author to write. It is also the last point in the chain at which anyone was careful. The range narrowed to its upper bound, the qualification fell away, and the subject widened from business process reengineering to organisational change in general. By the time the figure reached the slide decks it had become a measured failure rate for a category of activity its originators had not been describing.
A second commonly cited origin is John Kotter's 1995 Harvard Business Review article on why transformation efforts fail, which drew on his observation of more than a hundred companies over the preceding decade. Observation of a hundred companies is a real basis for an argument about what goes wrong. It is not a sampling frame, it does not yield a percentage, and Kotter's article does not present one in the way the citation trail implies.
The definitional problem is worse than the sourcing problem
Suppose a study existed. It would still have to answer three questions that no version of the claim answers.
What is a change programme? A merger, a new CRM rollout, a restructure of one department and a rebrand are all organisational change. They have nothing in common in duration, cost, reversibility or how anyone would know whether they worked.
What counts as failure? Abandoned before completion, delivered late, delivered on time but with no measurable benefit, delivered with benefits that decayed within a year: these are four different outcomes and the claim does not distinguish them. In practice most programmes deliver some of what was intended and not the rest, which makes a binary success rate the wrong shape of measurement entirely.
Measured when? A reorganisation judged at six months and the same reorganisation judged at three years will not get the same verdict, and no version of the claim specifies a horizon.
A statistic that cannot survive those three questions is not a weak statistic. It is a rhetorical device wearing a number.
Why it persists
The claim is useful to the people who repeat it most, and that is worth saying plainly rather than implying.
A high failure rate is the premise of the change management industry. If most change succeeds, methodology and tooling are optional. If most change fails, they are urgent. The figure is quoted in the first minute of a sales conversation because it converts a discretionary purchase into a risk mitigation, and it is quoted in the introduction of academic papers because a striking number establishes that the topic matters.
The scholarly literature has not been much better behaved. Stouten, Rousseau and De Cremer opened their 2018 review in the Academy of Management Annals by noting that organisational research lacks an easily accessible consensus on basic change management, which is a polite description of a field where the best known number is an unsourced one.
What to say instead
The absence of a real failure rate does not mean change is easy, and the correction should not be overcorrected into a claim that most change works. Nobody knows the rate, because nobody has defined the terms well enough to measure it.
What can be said is narrower and more useful:
- Specific, well-defined categories of change have been measured. Post-merger performance, IT project delivery and restructuring outcomes all have their own literatures with their own definitions, and those are the numbers worth citing when they fit the case at hand.
- The practitioner models used to manage change have very different evidential bases, and the differences are large. That comparison is in which change management model has evidence behind it.
- Where an intervention has been tested, the effects on leaders themselves are measurable and reasonably well established. What leadership training changes rests on 335 samples rather than on an estimate its own authors disclaimed.
If a supplier opens with the 70 percent figure, the useful response is to ask which study, and then to ask what that study counted as failure. The answer to the first question is that there isn't one. There has not been one for thirty years.
Common questions
Is it true that 70 percent of change initiatives fail? There is no evidence for it. A 2011 peer-reviewed review in the Journal of Change Management traced five published instances of the claim and found no valid and reliable empirical evidence behind any of them.
Where did the 70 percent figure originate? The earliest version usually pointed to is Hammer and Champy's 1993 book on reengineering, which gave a range of 50 to 70 percent for reengineering efforts that did not achieve intended results and described it as an unscientific estimate.
Did John Kotter say 70 percent of change fails? Kotter's 1995 Harvard Business Review article discussed why transformation efforts fail, based on observing more than a hundred companies over ten years. That is qualitative observation rather than a sampling study, and it does not produce a failure rate.
Has anyone measured the real failure rate of organisational change? Not for organisational change as a whole, because the term covers activities with nothing measurable in common. Narrower categories such as IT project delivery or post-merger performance do have measured outcomes.
Why do consultancies still use the figure? Because a high failure rate is the commercial premise for change management services. If most change succeeded without help, the methodology would be optional.
What should I say when someone quotes it to me? Ask which study it comes from, and what that study counted as a failure. Neither question has an answer, and asking them is faster than arguing about the number.
Does this mean change management is unnecessary? No. It means the case for it has to be made from evidence about specific practices rather than from an unsourced aggregate failure rate. Several individual practices do have supporting evidence; the headline statistic does not.
Sources
- Hughes, M. (2011), Do 70 Per Cent of All Organizational Change Initiatives Really Fail?, Journal of Change Management 11(4), 451-464
- University of Brighton research repository record for the paper, including the full abstract
- Stouten, Rousseau & De Cremer (2018), Successful Organizational Change: Integrating the Management Practice and Scholarly Literatures, Academy of Management Annals 12(2), 752-788
