Fidelity to design vs. fidelity to best practice: Lessons from our process evaluation of Benin’s curriculum reform

“Only 12 -13% of education program evaluations include any implementation data at all,” noted our Chief Technical and Learning Officer, Marc Shotland, in his reflections blog after this year’s Comparative and International Education Society (CIES) conference. Without implementation data, we rarely know how educational interventions are delivered on the ground, whether they succeed or fail, and whether fidelity is maintained or broken. The field, as Marc puts it, is “flying blind on the mechanisms.”

Central to that blindness is a distinction that What Works Hub for Global Education surfaced at their CIES workshop on implementation measurement: the difference between fidelity to design and fidelity to best practice. To explain:

In fidelity to design, a program can be implemented exactly as designed and still fall far short of what the evidence says works. Conversely, in fidelity to best practice, a frontline implementer might deviate from the written protocol in ways that actually bring them closer to best practice. 

We observed this distinction firsthand during our process evaluation of Benin’s early-grade curriculum reform, conducted in collaboration with the Institut National pour la Formation et la Recherche en Education (INFRE). The reform was being implemented as expected, but implementation alone did not guarantee effective teaching.

The reform 

In recent years, Benin’s Ministry of Preschool and Primary Education (MEMP) has led an ambitious curriculum reform focused on strengthening foundational literacy and numeracy. Central to the reform is a structured approach to instruction known as the gradual release model, commonly described as ‘I do, we do, you do.’ The logic is:

A teacher first models a skill explicitly (‘I do’), then practices it together with students (‘we do’), and finally gives students space to apply it independently (‘you do’). 

Each phase progressively shifts cognitive responsibility from teacher to student and creates a feedback loop. 

Alongside this pedagogical shift, the reform introduced new teaching and learning materials for CI and CP (roughly equivalent to kindergarten and Grade 1).

Our evaluation

With support from Engeza, a technical assistance fund for foundational learning financed by the Gates Foundation and managed by the Education Development Trust, IDinsight conducted a nationally representative process evaluation to assess how the reform was being implemented in classrooms. 

We observed over 100 reading and math lessons, interviewed teachers and school directors across 55 schools, and reviewed instructional materials. Our goal was to document how the reform’s pedagogical principles were translating into classroom practice and to identify concrete opportunities to strengthen its implementation. A summary of our findings is available here.

Our observation: High adoption, variable depth

By conventional implementation metrics, the education reforms in Benin have made significant progress. Teachers have adopted the gradual release model. In most classrooms we observed, lessons followed the intended sequence: the teacher modelled a skill, practised it with students, then gave students opportunities to practice independently. The new materials, including teacher guides and student resources, were actively in use by almost all teachers. 

This represents an important achievement: the reform’s core design has been effectively communicated and is being applied at scale. By the measure of fidelity to design, the reform was succeeding.

At the same time, adoption of the structure does not automatically translate into the responsive, interactive teaching that makes the gradual release model effective. In many classrooms, lessons followed the expected sequence but offered limited opportunities for genuine student engagement. Teachers moved through the phases (‘I do, we do, you do’), but their adaptation of instruction based on student responses was inconsistent. Formative assessment occurred structurally; teachers checked answers, circulated, and managed errors, but the information rarely fed back into instructional adjustments. The ‘we do’ and ‘you do’ phases, where students practice with teacher support, often relied on repetition and copying rather than activities that pushed students to actively mobilise what they had learned.

The structure was in place. But fidelity to best practice was observed less consistently.

Why this gap matters

The gradual release model of teaching works because it creates feedback loops. The teacher models, students practice with support, the teacher observes what students are doing, and instruction is adjusted in real time. Without that adjustment, the approach risks becoming a script rather than a pedagogy. Evidence shows that the interactive and adaptive elements are essential to effective explicit teaching (Archer & Hughes, 2011)

This distinction between fidelity to design and fidelity to best practice has practical implications for how we evaluate reforms. Asking “Are teachers using the method?” is not the same as asking “Are teachers teaching effectively?” A process evaluation that stops at adoption metrics, such as lesson structure followed, materials used, and training completed, can report success while missing the gap that explains why learning outcomes may not yet reflect the investments made.

What drives this gap

Our evaluation pointed to three factors:

1. Learning materials matter, but they are not sufficient on their own. The new instructional materials supported teachers in structuring lessons but offered limited guidance on incorporating responsiveness in classrooms. In reading materials, emphasis was placed on decoding but provided less support for vocabulary development, comprehension, and oral language, areas that are particularly important for students learning in French, which is generally not their home language. In math, materials focused on procedural knowledge with fewer opportunities to develop conceptual understanding.
2. Training shapes practice, but coverage was uneven. In math, stronger implementation was found in classrooms where the teacher had received targeted training on the new CI and CP methods. The ministry delivered this training to the first cohorts, with school directors expected to extend it to teachers who had not been reached directly. Closing the gap means deeper training focused on pedagogical agility: remediation, teaching in a second language, and teaching for conceptual understanding in math. It also means building that mastery first among the inspectors, advisors, and directors who support teachers, through coaching and analysis of their own practice, and offering continuing education through e-learning and video.
3. The middle tier is underleveraged. Pedagogical advisors, inspectors, and school directors sit between policy and classrooms, and our findings reinforced how consequential that layer is. In our analysis, post-observation feedback from school directors was associated with stronger classroom practices. Yet this tier lacks the practical tools and dedicated time to guide their feedback and follow-up. Without that support, their classroom visits become compliance checks rather than opportunities for instructional improvement.

Recommendations and Implications for the field

We discussed four recommendations with the Ministry, structured around short-term actions and medium-term orientations:

• Strengthen materials to support interactive practices. This includes developing supplementary reading activity guides that support vocabulary, comprehension, and oral language, as well as concept-focused math posters that help teachers move beyond procedures to understanding.
• Improve coherence between curricula, materials, and training. Clarify expectations around lesson pacing and the role of each phase of explicit teaching. Teachers benefit from understanding not just the sequence but the purpose of each step, and what responsive practice looks like within it.
• Adapt guidance for constrained contexts. Large class sizes are a reality in many Beninese schools. Implementation support should address how to maintain interactive practices in large classes.
• Equip the middle tier with practical feedback tools. Pedagogical advisors and school directors would benefit from simple, usable instruments to guide classroom observations and follow-up conversations, with a focus on instructional quality rather than only on compliance with structure.

Looking ahead

While our recommendations are specific to Benin, we recognise that the underlying tensions are common across structured pedagogy reforms in low- and middle-income contexts. Teams designing or evaluating similar reforms elsewhere are likely to encounter the same challenge of moving from adoption to quality. 

Fidelity to design is achievable; Benin demonstrates that. But fidelity to best practice is harder, slower, and less visible in the data our sector typically collects. Closing that gap requires understanding why teachers struggle to be responsive in practice, what would help them shift, and giving them the tools and support to do so. That work begins with measuring implementation.