You will have heard it. It appears on conference slides, in consultancy proposals, in the opening paragraph of more or less every article about organisational change. Why change management fails is usually presented as a question with a simple answer: seventy per cent of change programmes fail.
It’s a compelling number. It’s also, on the available evidence, unsupported.
The most-quoted statistic in change management has never been reliably demonstrated.
Where the figure came from
The claim is generally traced to the business process reengineering literature of the early 1990s, where it appeared as an observation rather than a research finding — and was then repeated, cited, and re-cited until repetition made it feel established.
In 2011, Mark Hughes examined it directly in the Journal of Change Management in a paper titled “Do 70 Per Cent of All Organizational Change Initiatives Really Fail?” His conclusion was blunt: whilst a popular narrative of 70% failure clearly exists, “there is no valid and reliable empirical evidence to support such a narrative.”
Others have made the same point since. The figure persists anyway, largely because it’s useful — it makes change sound difficult, which suits anyone selling help with it.
Why we’re bothering to correct it
Two reasons, and neither is pedantry.
First, it’s bad for the businesses that hear it. A leadership team told that change fails 70% of the time reasonably concludes that change is a lottery. That’s demoralising and it isn’t true. It also provides advance cover for failure — if most programmes fail, ours failing was statistically normal rather than something we could have prevented.
Second, it obscures the actual question. “Does change fail?” is unanswerable without defining what counts as change and what counts as failure. A far more useful question is: what distinguishes the changes that hold from the ones that quietly unwind?
Asking why change management fails in a specific organisation, rather than in general, produces answers you can act on.
What we actually see
Across improvement work in construction, housing and manufacturing, the changes that survive tend to share a small number of characteristics. The ones that don’t survive tend to fail in recognisable ways.
Change slips back when nobody owns it. This is the most common failure by some distance. A new process is designed, implemented, and then belongs to no one in particular. Processes drift by default — under pressure, people revert to what they know. Someone has to be responsible for noticing.
Change slips back when the old way still works. If the previous route remains available, some proportion of people will use it, particularly when busy. Parallel systems are the enemy of adopted change. If the new process is genuinely better, the old one needs closing rather than deprecating.
Change slips back when it was never measured. Without a baseline, nobody can tell whether the change delivered anything. It becomes a matter of opinion, and opinion decays. The RICS Construction Productivity Report 2026 found 22% of UK construction firms never measure productivity — you cannot sustain an improvement you cannot detect.
Change slips back when it was designed without the people doing the work. They know where the process actually fails. If they weren’t asked, the redesign usually contains an error they could have prevented, and they have no investment in making it work.
Change slips back when the incentives contradict it. If the new process is safer but slower, and people are still measured purely on speed, the process will lose. This one is nearly always invisible to the people who designed the change.
What makes it hold
Reverse the above and you have most of the answer.
- A named owner who is accountable for the process still running correctly in six months, not for implementing it
- The old route closed, not merely discouraged
- A baseline and a follow-up measurement, so the effect is a fact rather than a belief
- Genuine involvement from the people who run the process daily, at design stage rather than rollout
- Aligned incentives, so nobody is punished for doing the new thing properly
- Visible leadership behaviour — managers using the new process themselves, including when it’s inconvenient
Change doesn’t fail because people resist it. It fails because organisations design it and then stop paying attention.
The reinforcement gap
There’s one more pattern worth naming, because it’s the one businesses find hardest to accept.
Most change effort is front-loaded. The launch gets the communication, the training, the leadership attention. Then attention moves to the next thing.
But the point at which change is most fragile is not the launch. It’s roughly three months in — when the initial enthusiasm has gone, the first proper crisis has arrived, and reverting to the old way is the fastest route through it.
Almost nobody plans for month three. It is the cheapest possible intervention and the one most often missed.
What this means for training
Where training fits into this is worth being precise about.
Training at launch teaches people the new process. That’s necessary and insufficient. What sustains change is capability in the people who manage it — managers who know how to hold a standard, address drift early, and have the conversation when someone quietly reverts.
That’s why we tend to build continuous improvement capability into a business rather than delivering a project and leaving. A team that can run its own improvement cycles doesn’t need the change to be permanent, because it can keep changing.
Our post on embedding and sustaining change in business → /embedding-and-sustaining-change goes further into the practical mechanics.
What to do about the number when you hear it
A practical suggestion, since the figure is not going away.
When somebody quotes it at you — in a proposal, a pitch or a board paper — ask two questions. Which study? And what counted as failure?
Neither is a hostile question. Both are reasonable. And the answer is almost always that the person quoting it doesn’t know, because it arrived as received wisdom rather than research.
That exchange is worth having, because it changes the conversation from “change is inherently risky” to “what specifically would make this one work”. The first framing invites fatalism. The second invites planning.
Why this matters for how you buy help
There is a commercial point buried in this too.
A supplier who opens with a failure statistic is establishing, before any work begins, that a poor outcome would be normal. That is a useful position for them and a poor one for you.
Better questions to ask anybody proposing change work: who will own this in six months, how will we know whether it held, and what happens at month three when the initial energy has gone. A supplier with good answers to those is worth more than one with a compelling slide about failure rates.
The honest version
Change is difficult. Plenty of programmes deliver less than they promised. But the idea that failure is the statistical norm is a claim without evidence, and businesses would be better served by dropping it.
What determines the outcome is mostly unglamorous: ownership, measurement, involvement, incentives, and paying attention for longer than the launch.
None of that requires a 70% failure rate to justify. It just requires doing the work.
Explore our business training workshops → /training-workshops , or talk to us → /contact about building improvement capability in your team.