Seventy Per Cent.
For thirty years, one figure has opened almost every transformation programme. We went looking for where it came from.
Seventy per cent of corporate change efforts fail. You have seen the figure, and so has every executive who has ever commissioned a transformation. It opens the pitch deck and the board paper, and it justifies the budget.
In 2011 a management researcher named Mark Hughes went looking for the study underneath it. There wasn't one.
He followed the citations backwards. Today's digital-transformation reports cited the business books of the 2000s, which cited the reengineering manifestos of the 1990s. Every footnote pointed to another footnote, never to a measurement.
The trail ended at one book: Reengineering the Corporation, by Michael Hammer and James Champy, 1993. Hughes turned to the page expecting a study. He found a sentence.
Hammer and Champy had written that, by their own unscientific estimate, somewhere between fifty and seventy per cent of reengineering efforts did not achieve the results their sponsors intended. They flagged it as a guess, and added that it did not make reengineering risky. The higher end of their guess was the part that travelled.
There was no survey behind it and no dataset anyone could re-open. The number that has governed thirty years of corporate spending was a working impression about a movement then four years old.
Hughes published the trace. The field has argued over it ever since, and in more than a decade no one has produced the missing study.
Even without a study beneath it, the underdelivery the number points to is real. Programmes do fall short of what their sponsors set, and the rate has held steady for thirty years. Folklore and evidence agree on that much. Where they part company is the question the folklore never asked: why the programmes fall short.
In 2008 McKinsey put it to more than three thousand executives who had just run major transformations. About a third called their programme a clear success. So far, the familiar number.
Then the survey asked the other two-thirds what went wrong. The answers pointed to executive teams that were not aligned, ambitions scoped beyond what the organisation could absorb, and a shortfall in the frontline skills the new model had assumed it would have. By the executives' own account, the failure sat mostly above and around the front line.
Longitudinal audits from Prosci and Deloitte came at it from the shop floor. Ask frontline staff why a programme stalled, and the top answer is that the new operating model made their core daily work harder to do, sometimes impossible.
The figures are theirs. Even they locate it in the architecture.
The data says one thing and the playbooks say another. The data locates the shortfall in the design — alignment, scope, whether the new model let people do the work. The playbooks locate it in the person.
Kotter opens with urgency and a guiding coalition: the obstacle is reluctance, the cure is belief. Prosci's ADKAR walks each individual from awareness through to reinforcement. Each takes one thing as settled — that the new model is sound, and the only variable left to manage is whether people accept it.
That assumption is the floor they stand on. It is also the spot where they run out of room, because neither has anywhere to put a finding that says the model itself was what made the work harder.
Which leaves a question thirty years of better methods have not answered. All that effort went into the workforce's willingness to change. If the number measured willingness, it should have moved. It hasn't.
So what if the durable part of the seventy per cent was never about people at all — but about something they were responding to?
Lift the story about human nature off the number, and the number changes what it points at. It stops describing the workforce and starts describing a distance — between where the work is designed and where it is done.
Everyone read the figure as a fever in the patient. It was the temperature of the building.
A transformation is authored at the centre: a programme office, a steering committee, a vendor's reference architecture, a slide a few senior people sign. Then it is handed to an edge that did not help write it. The further those two points sit apart, the more of the design dies on contact with the actual work. Seventy per cent is the size of that gap, averaged across a generation of programmes. It held steady because the distance held steady.
What gets called resistance is often a workforce protecting its ability to finish the day. It steps around an installation that makes the day's work harder. The dashboard flags this as non-compliance; it is closer to an immune response — a working system rejecting a payload that threatens its function.
The story blaming the worker was useful — it kept the architecture off trial. The framework was sound and the strategy was right; the project simply met the immovable fact of people. An industry that was never paid against outcomes had little reason to look past an explanation that pointed so reliably away from its own work.
For most of its life the gap was an expensive inefficiency a company could carry, because both sides moved at human speed. The design was slow to write and the execution slow to run, and the friction between them sat inside a tolerance everyone had budgeted for.
That symmetry is now ending.
Someone at the edge can now produce in three minutes, with a generative model, what used to take three weeks. The binding constraint is no longer their willingness to use the tool. They are already using it, often before the centre approved it.
The constraint is whether the architecture will let the output move — the approval chain built for documents that arrived weekly, a sign-off that assumes a week of preparation now done in an afternoon, controls timed to the speed of a human hand and now meeting the speed of a model.
The first reading of a stalled AI programme asks how to raise adoption — more communication, more training, more dashboards. It is the answer the seventy per cent has always been given. The number has not moved.
The second asks where the structure strips people of the licence to use what they have already adopted. Which approval was built for documents that no longer arrive. Which sign-off assumes a week now done in an afternoon.
The organisations pulling ahead are the ones who noticed the bottleneck had moved. They stopped spending against the workforce's willingness and started clearing the choke points between a capable person and a finished piece of work. Their number begins to fall, because the distance closed.
Seventy per cent measured how far the work has drifted from the people who do it. For thirty years the figure sat still, because the distance it measured sat still. That distance is now closing.
The final section — the reframe — is reserved for members. Enter your email to continue reading.