What the research found
Mark Hughes’s 2011 study for the Journal of Change Management, “Do 70 Per Cent of All Organizational Change Initiatives Really Fail?”, offers a directly relevant case study in how a widely repeated benchmark can fail scrutiny. Hughes examined five separate published instances of the commonly cited claim that seventy per cent of organisational change initiatives fail, and found no valid or reliable empirical evidence supporting any of them. The figure’s earliest traceable origin sits with Hammer and Champy’s 1993 work, where it was presented explicitly as an “unscientific estimate” — a characterisation the figure’s subsequent decades of repetition have almost entirely stripped away. A separate, methodologically transparent McKinsey survey of 1,546 executives found roughly 30 per cent reporting change programmes as completely or mostly successful and only around 10 per cent as unsuccessful — numbers that do not resemble a 70 per cent failure rate in any interpretation.
Paul Rogers and Marcia W. Blenko’s Bain-sponsored research, published in Harvard Business Review, offers a complementary but distinct angle: even where a benchmark figure is methodologically sound, its usefulness depends heavily on whether the underlying comparison — the population surveyed, the definitions used, the operating context assumed — genuinely matches the organisation applying it. Their finding that only around 15% of surveyed executives reported having decision-making structures that helped their business outperform competitors is itself a benchmark of sorts, and one whose value depends on a leadership team understanding what population that 15% figure actually describes, rather than treating it as a universal constant.
V. F. Ridgway’s foundational research on performance measurement, published in Administrative Science Quarterly, contributes a further and more general caution relevant to any benchmark built from a single quantitative comparison: numbers used indiscriminately, without attention to how they were produced and what they can and cannot capture, tend to produce exactly the kind of overconfidence that makes a misleading benchmark dangerous rather than merely imprecise.
Where it converges and where it diverges
Read together, these bodies of research converge on a single underlying theme: the risk in benchmarking rarely lies in the practice of comparison itself, which is sound in principle, but in the gap between how a benchmark figure is actually produced and how confidently it is subsequently used. Hughes’s research demonstrates this gap at its most severe — a figure with no traceable empirical basis at all, repeated for decades regardless. Rogers and Blenko’s research demonstrates a subtler version — a figure that is methodologically sound but easily misapplied outside the specific population it describes. Ridgway’s research supplies the general mechanism connecting both cases: quantitative comparison, trusted indiscriminately, produces exactly this kind of overconfidence.
Where the research diverges is on where the responsibility for catching the problem should sit. Hughes’s work implicitly places responsibility on whoever first repeats a figure to trace its origin before using it — a high bar that, in practice, is rarely met, which is precisely why an untraceable figure like the 70 per cent claim can survive for decades inside organisations that would never accept an equally unsupported internal number. Rogers and Blenko’s research, by contrast, suggests the more common failure is not a fabricated figure but a genuine one applied to the wrong context — which places the burden less on tracing origin and more on checking comparability: does the benchmark population actually resemble this organisation, in operating model, scale and market, closely enough for the comparison to hold?
Not every benchmark is unsound, and it is worth stating plainly that this body of research does not support blanket scepticism toward external comparison. A benchmark whose group is genuinely comparable, and whose figure is traceable to a transparent methodology, can be a legitimate and valuable check on internal performance. The research collectively argues for treating each specific benchmark as something to be verified on its own terms, rather than trusting or dismissing benchmarking as a category.
A further point of convergence concerns why misleading benchmarks tend to be especially persuasive to leadership teams specifically, as distinct from any other audience inside or outside an organisation. A benchmark carries an external, seemingly objective authority that an internal number does not — it appears to represent a market reality rather than an internal opinion, which makes it harder to challenge in the room where it is presented, particularly if no one present has independently traced its origin. This dynamic is consistent with Ridgway’s broader caution about indiscriminate trust in quantitative comparison: the more a number appears to come from outside the organisation, the less scrutiny it tends to receive from within it, even though its outside origin says nothing, on its own, about whether the number is accurate or applicable.
This also helps explain why a debunked or untraceable benchmark, once it has entered a leadership team’s regular vocabulary, can be unusually difficult to dislodge even after being challenged. Hughes’s research found the 70 per cent change-failure claim persisting in circulation for decades after its original “unscientific estimate” framing had been dropped, and after its author had publicly walked it back. A figure that has become part of how a leadership team routinely frames a decision — used in slide decks, cited in planning conversations, treated as a shared point of reference — has a kind of institutional inertia that an individual correction rarely overcomes on its own, which is a separate and additional reason misleading benchmarks are worth checking before they enter that kind of routine use, rather than after.
There is a further, quieter risk worth naming: a benchmark can mislead not by being wrong, but by narrowing what a leadership team pays attention to. Once a comparison to a named competitor or industry figure enters the conversation, it tends to organise the discussion that follows around beating, matching, or explaining a gap to that specific reference point — even where the reference point, however accurate, is not actually the most relevant comparison available for the decision at hand. An organisation genuinely ahead of a flawed or narrow benchmark can end up celebrating a comparison that tells it very little about whether it is actually performing well, simply because the benchmark, once introduced, became the frame the whole conversation was built around.
What remains unresolved
How widely untraceable but persistently repeated figures — comparable to the 70 per cent change-failure claim — continue to circulate inside leadership teams without being checked, and what makes a figure of this kind especially durable once it enters common use.
Whether leadership teams are, on balance, more vulnerable to a benchmark that is simply wrong, or to one that is technically accurate but drawn from a population too different from their own to apply safely — the research examined here suggests both failure modes are real without offering a clear answer on which is more common in practice.
What proportion of benchmarks currently in active use inside a typical organisation have ever actually been traced back to their original source, as opposed to being accepted because they were already in circulation when they arrived.
For how this pattern presents in other kinds of organisation, see other sectors.
Questions worth considering
- For each benchmark currently in use, has anyone traced the figure back to how it was produced and what it can and cannot capture?
- Is the organisation comparing itself with firms that are genuinely like it, or with whichever figure happened to be in circulation?
- Once a benchmark has framed the conversation, who would feel able to question it?
More on this topic: Performance.
Your team is facing similar challenges?
Subscribe to The Riddle Ledger
Perspectives are sent as they are published, with no promotional email, and you can unsubscribe at any time.