When should a pilot scale?

Pilots are for learning, not proving

A pilot often begins with a quiet political pressure: prove the idea works. That pressure encourages teams to protect the intervention, explain away problems and treat positive stories as confirmation.

A useful pilot has a different job. It reduces uncertainty. It tells decision-makers which elements create value, which assumptions were wrong, what the model costs under real conditions and what must be true before more people or places are added.

That means a pilot can be valuable even when the original model does not survive unchanged.

The purpose of a pilot is not to defend the idea. It is to improve the decision.

Separate the result from the conditions

Strong outcomes may be real and still not be scalable. A small pilot often benefits from unusually committed leaders, hand-picked schools, intensive coaching, generous equipment, short communication lines and the constant attention of the design team.

Before scaling, ask which conditions were essential. Could a typical school reproduce them? Can the available workforce provide the same support? Can procurement, safeguarding, data quality and maintenance survive greater distance and volume?

• People: Were results dependent on exceptional individuals?
• Process: Is delivery documented, teachable and usable?
• Infrastructure: What minimum conditions are genuinely necessary?
• Economics: What changes when volume, travel and supervision increase?
• Governance: Who decides, monitors and responds when delivery varies?

Five gates for a responsible scale decision

A decision to scale should pass through explicit gates. Not every programme needs perfect evidence, but every programme should be clear about the uncertainty it is accepting.

• Value: Participants experienced a meaningful improvement, not merely activity.
• Delivery: The core model was implemented with acceptable quality across more than one setting.
• Capability: People within the system can carry the work with proportionate external support.
• Economics: Costs are understood, affordable and linked to the value created.
• Adaptability: The model has clear non-negotiables and room for context-sensitive adaptation.

Scale the learning system before the intervention

At ten sites, a programme lead can personally notice drift. At one hundred sites, the model needs an early-warning system. Scale requires a cadence for data, observation, escalation, reflection and change.

This does not mean collecting more data. It means collecting the few signals that allow the system to act: participation, quality, capability, risk, outcome and cost. A dashboard without a decision rhythm is decoration. A monthly learning conversation with clear owners can be a management system.

The ability to detect and correct variation is part of the programme—not an administrative extra.

There are more choices than scale or stop

A good pilot may lead to expansion, but it may also lead to another cycle of validation, a narrower target group, a different delivery partner, lower-cost infrastructure, a revised learning sequence or a cluster model instead of full replication.

Sometimes the responsible decision is to pause. That is not failure. It protects public trust, beneficiary experience and future investment from a model that has not earned confidence.

Scale when evidence, capability and economics point in the same direction. Until then, keep learning deliberately.

• Scale now.
• Adapt and test again.
• Narrow the model.
• Strengthen system readiness first.
• Stop and redirect resources.