Why promising interventions can produce different results at scale and how governments can use pilot evidence to strengthen nationwide implementation?

Pilot studies help governments test whether an intervention can work before committing public resources at scale. But a successful pilot answers only one question: can this intervention work under the conditions in which it was tested? Scaling asks a harder one: can it continue to work when delivered across different geographies, institutions, frontline workforces and beneficiary groups?

India’s experience across education, nutrition and renewable energy demonstrates why this distinction matters. Teaching at the Right Level (TaRL), rice fortification and PM-KUSUM each offer different lessons about moving from controlled or limited implementation towards larger public programmes.

The evidence does not suggest that pilots or subsequent programmes should simply be classified as successes or failures. Instead, it shows that scale changes the conditions under which a policy operates. Implementation fidelity can weaken; institutional incentives can change; supply chains become more complex; beneficiaries become more heterogeneous; and a pilot may not have tested every outcome that becomes important at scale.

The policy lesson is therefore not to pilot more cautiously, but to design pilots for scale, evaluate what changes during expansion, and treat scaling itself as a continuing process of learning and adaptation.

1. Why This Matters

Governments routinely use pilots to reduce uncertainty before committing substantial fiscal and administrative resources. A pilot can test an intervention’s feasibility, reveal implementation difficulties and provide evidence for redesign.

But moving from 100 schools to thousands, one district to hundreds, or a handful of farmers to millions is not simply multiplication.

The operating environment changes.

A pilot may have a dedicated implementation team, intensive monitoring, motivated frontline workers, selected locations and close technical support. At scale, the same intervention may depend on regular government staff, multiple administrative levels, different state capacities, large procurement systems and beneficiaries with widely varying circumstances.

This creates what implementation research often describes as “voltage drop”: the effect observed under pilot conditions can become smaller when an intervention enters routine systems.

The appropriate policy question is therefore not:

Did the pilot succeed?

It is:

Which elements made the pilot work, and can those elements survive at scale?

Three Indian experiences illustrate three different answers.

2. Three Indian Lessons from Moving to Scale

Case 1: Teaching at the Right Level — Preserve the Mechanism, Not Just the Programme

Pratham’s Teaching at the Right Level (TaRL) approach begins with a simple observation: children in the same grade can have very different learning levels. Instead of teaching exclusively according to grade-level curriculum, children are periodically assessed, grouped according to learning level and given targeted instruction.

Earlier evaluations of Pratham programmes demonstrated substantial learning gains. The Balsakhi programme, evaluated in government primary schools in Vadodara and Mumbai, showed that targeted remedial instruction could improve learning, particularly among children who had fallen furthest behind.

The challenge emerged when the underlying approach was introduced into ordinary government-school systems.

Early scale-up efforts in Bihar and Uttarakhand found that training teachers and supplying materials were not enough. According to subsequent research by Banerjee and colleagues, teachers generally did not adopt the methodology as intended despite training and Pratham support.

This revealed an important distinction:

The intervention was not merely “teacher training.” Its mechanism depended on teachers actually reorganising classroom time around children’s learning levels.

Researchers and implementers subsequently adapted the model.

In Haryana, teachers received support from government resource persons and were given a dedicated period in which to implement the approach. In Uttar Pradesh, intensive short-duration learning camps were delivered during school hours. These redesigned models generated measurable learning gains: approximately 0.15 standard deviations in language in Haryana and 0.70 standard deviations in Uttar Pradesh in the evaluated models.

Lesson for policymakers

This case should therefore not be presented as an example of a pilot that “failed at scale.”

It demonstrates something more useful:

When scale reveals that the original delivery mechanism does not fit routine administrative conditions, implementation design must be adapted without losing the intervention’s core mechanism.

Training, materials and instructions are inputs. The real question is whether the behaviour that generated the original result survives.

Case 2: Rice Fortification — A Pilot Must Test the Questions That Matter at Scale

India’s rice-fortification initiative addresses micronutrient deficiencies by blending fortified rice kernels containing iron, folic acid and vitamin B12 with regular rice.

Its potential reach is enormous because rice is distributed through large public programmes.

The Government of India approved a three-year Centrally Sponsored Pilot Scheme beginning in 2019–20, with a budget of ₹174.64 crore and a focus on 15 districts across participating states. By February 2022, 11 states had reported distribution of 3.64 lakh metric tonnes of fortified rice through their identified pilot districts.

The initiative subsequently expanded in phases. The Union Cabinet approved nationwide expansion through TPDS and other welfare schemes, and in October 2024 approved continuation of fortified-rice supply through December 2028.

The policy lesson here differs from TaRL.

A pilot can successfully answer questions about production, blending, procurement and distribution without necessarily answering every question that becomes important when coverage expands dramatically.

At national scale, nutritional outcomes, quality assurance, beneficiary heterogeneity, monitoring and safety considerations become increasingly important alongside logistical feasibility.

There has also been debate over the health evidence and safeguards. Government sources state that scientific evidence supports the safety of fortified rice, including for people with haemoglobinopathies, while critics and some public-health researchers have raised questions about mandatory fortification, population-specific risks and the strength of evidence for reducing anaemia. These are competing interpretations that should be presented as such rather than resolved editorially.

Lesson for policymakers

The scope of evaluation should expand with the scale and consequences of the intervention.

Before moving from district-level implementation to population-scale delivery, policymakers should ask whether the pilot tested:

• Can we deliver it?
• Does it achieve the intended outcome?
• For whom does it work?
• Are there groups requiring additional safeguards?
• Can quality remain consistent at scale?

Operational feasibility is important but it is not always sufficient evidence for every subsequent policy decision.

Case 3: PM-KUSUM — Preserve Incentives When Scaling Technology

PM-KUSUM illustrates a third scaling challenge: technology can be easier to replicate than the incentives surrounding it.

The scheme supports decentralised solar generation and agricultural solarisation through three components. Its stated target is 34,800 MW of solar capacity with central financial support of ₹34,422 crore.

One important precursor was the Dhundi Solar Power Irrigators’ Cooperative Enterprise in Gujarat. Farmers using solar pumps could sell surplus electricity, creating an economic reason not to use every available unit of solar power for groundwater pumping.

This matters because solar irrigation changes the economics of water extraction. Once installed, a standalone solar pump has very low marginal energy cost. Without complementary groundwater management or incentives to conserve electricity and water, cheap energy can potentially encourage greater pumping.

PM-KUSUM’s components, however, serve different purposes and should not be evaluated as if they were one intervention. Component B supports standalone solar pumps in off-grid areas. Component C supports solarisation of grid-connected pumps and allows surplus electricity to be sold to DISCOMs.

Implementation performance has also differed substantially between components. As of 31 July 2026, the official PM-KUSUM dashboard reported 1,432.92 MW installed under Component A against 7,214 MW sanctioned; 7.26 lakh standalone pumps installed under Component B against 8.12 lakh sanctioned; 8,444 individual grid-connected pumps solarised under Component C-IPS; and more than 8.59 lakh pumps covered through feeder-level solarisation.

The important analytical point is therefore not that PM-KUSUM “discarded” the Dhundi model. The national scheme is broader and serves several policy objectives.

The lesson is that different scaling pathways preserve different parts of the original incentive architecture.

Lesson for policymakers

When a pilot works because technology and incentives reinforce each other, scaling the technology alone may not reproduce the same behavioural outcome.

For interventions involving groundwater, electricity, common resources or behavioural incentives, policymakers should therefore evaluate not only installation numbers but also how the scaled design changes user behaviour.

3. What Actually Changes Between Pilot and Scale?

The three cases reveal that scaling problems do not have a single cause.

This produces a fundamental principle:

Scale is not a larger pilot. It is a different operating environment.

4. What India Can Learn

India’s federal structure makes this particularly important.

A programme that works in one district may enter states with different administrative capacity, infrastructure, fiscal space, geography and beneficiary behaviour. Uniform national objectives therefore often require non-uniform implementation pathways.

Three implications follow.

First, policymakers should distinguish between core and adaptable components. The mechanism responsible for the outcome must be protected; implementation details that depend on local conditions should be allowed to change.

Second, pilots should deliberately include difficult environments, not merely locations where implementation is easiest. A representative pilot portfolio may reveal more about scalability than one exceptionally successful demonstration site.

Third, evaluation should continue after scale-up begins. Haryana and Uttar Pradesh’s TaRL experience is instructive precisely because early difficulties generated redesign rather than a binary decision to either abandon or mechanically continue the original model.

The objective should therefore be adaptive scaling.

5. Tatvita Scale-Readiness Diagnostic

Before moving from pilot to large-scale implementation, decision-makers can test an intervention across six questions:

A pilot should therefore not receive a simple “successful/unsuccessful” label.

It should produce a scale-readiness assessment identifying what can be expanded immediately, what requires adaptation and what requires further evidence.

6. Tatvita SCALE Framework™

A more disciplined transition can follow five stages:

S — Specify the Mechanism

Identify exactly what produced the pilot outcome. Was it technology, incentives, staffing, supervision, financing—or their combination?

C — Check Context and Capacity

Test whether administrative capacity, beneficiary characteristics, infrastructure and institutional incentives differ across intended locations.

A — Adapt Before Expansion

Preserve the core mechanism while modifying delivery to fit routine systems.

L — Learn During Scale-Up

Use phased expansion, real-time monitoring and independent evaluation to detect declining effectiveness or emerging risks.

E — Evaluate Outcomes, Not Just Coverage

Installation, enrolment, distribution and expenditure measure implementation. They do not necessarily measure impact. Track the outcome the programme was created to improve.

7. Action Agenda: From Pilot Approval to Adaptive Scale-Up

India can strengthen pilot-to-scale transitions without slowing policy implementation.

Before the pilot, define the eventual scale-up pathway and identify the assumptions that need testing.

During the pilot, measure not only whether the intervention works, but why it works, what it costs and which institutional capabilities it requires.

Before expansion, separate the programme’s essential mechanism from features that can be locally adapted.

During early scale-up, expand across deliberately different environments and create feedback loops between frontline implementation and programme design.

At national or state scale, monitor outcomes alongside coverage, expenditure and physical targets.

Most importantly, build adaptation triggers into programme governance. If implementation data reveal that a mechanism is weakening, policymakers should be able to modify delivery without treating redesign as an admission of policy failure.

Key Takeaways

1. A successful pilot demonstrates potential—not automatic scalability. Scale changes institutions, populations, costs and implementation conditions.

2. Protect the mechanism, not necessarily the original delivery model. TaRL shows that redesign can improve the fit between an effective idea and routine government systems.

3. Test the outcomes that matter at the scale being contemplated. Operational feasibility, effectiveness, safety, affordability and equity are different questions.

4. Incentives matter as much as technology. PM-KUSUM illustrates why scaled programmes should examine how programme design affects behaviour, not only physical deployment.

5. Scale-up should remain an evidence-generating process. Expansion is not the end of experimentation; it is the next stage of learning.