Explore / Counterposition
“Successful pilots should be replicated faithfully”
Scale the load-bearing parts faithfully. Let the rest adapt, on purpose.
Published September 2026
The strongest formulation
A pilot that worked is evidence about a specific design. Changing it as it spreads throws away that evidence, and uncontrolled adaptation is how effective programmes are diluted to nothing.
Why it is attractive
It is simple to specify, fund and audit. Fidelity can be checked with a list.
Where it holds
It holds for the components that carry the effect. If the mechanism depends on a dose, a sequence or a feedback cycle, those should not drift.
Where it breaks
It breaks because pilots run in unusual conditions: volunteer staff, extra attention, novelty and support that will not scale. Practitioners adapt anyway, and a design that forbids adaptation gets adapted badly and invisibly. The disagreement is mostly one of evidence: which parts carried the effect.
What evidence matters
Berman and McLaughlin’s RAND Change Agent study (1978) on mutual adaptation. Elmore (1996) on getting to scale, and Coburn (2003) on scale as depth, sustainability, spread and ownership rather than numbers. What matters is which components carried the effect, and whether it survives outside the pilot’s conditions.
More precisely
A more precise version
A successful pilot scales when its load-bearing components are kept intact and the rest are adapted deliberately to each new setting.
Where next
Follow this idea
Operationalise it
Programme evaluation (opens in a new tab)Guide
Related argument
Design for translationNote
Related argument
Nobody in the room has to be wrongNote
Challenge it
WP-04: A scalable intervention must specify what may change as clearly as what must remain invariant.Working proposition