Reference Chapter 74 · Turning promising education ideas into dependable public service—and retiring those that do not work
Every Ministry of Education is asked to improve something. New mathematics teaching methods, assessment arrangements, teacher coaching programmes, digital platforms, language support systems and school schedules may promise better outcomes. A small project can appear successful when a dedicated principal and enthusiastic teachers invest extraordinary effort. The difficulty begins when the ministry tries to reproduce the project across hundreds or thousands of ordinary schools, each with different staffing, facilities, learners and capacity.
The gold standard is a national education reform implementation system that explores problems carefully, pilots proportionately, tests real learning, funds the conditions for scale, monitors unintended consequences and stops or redesigns programmes that cannot deliver their purpose. It should not confuse a positive pilot press release with credible evidence of national readiness. Equally, it should not freeze useful innovation behind endless approvals or demand every promising teaching idea undergo the same expensive evaluation.
This chapter is an operational acceptance manual for designing and scaling education reforms. Chapter 12 covers strategic foresight and policy renewal; Chapter 16 covers research and evidence synthesis; Chapter 45 covers the ministry’s operating model. This chapter connects them through a practical reform lifecycle with named gates, evidence standards, implementation resources and termination rules.
1. Define the problem before choosing the intervention
A minister may announce a new AI learning platform when the actual problem is a missing algebra prerequisite. A school may request longer lessons when its dominant failure is inaccurate teaching material. An expensive reform can be delivered flawlessly and still solve the wrong problem.
A National Education Problem Statement should describe:
- The affected learner group and educational stage.
- The intended capability or entitlement.
- Evidence that the current service is insufficient.
- The size, geography and distribution of the problem.
- Important competing explanations.
- Existing solutions or programmes.
- Constraints that limit current provision.
- The conditions under which improvement would be meaningful.
- The consequences of doing nothing or delaying action.
- The authorised public body responsible for deciding.
Do not define the problem as “schools do not have our proposed software.” That assumes the intended solution before the educational need has been established.
2. Use a staged but flexible implementation lifecycle
The Education Endowment Foundation’s 2024 School’s Guide to Implementation describes four flexible phases: Explore, Prepare, Deliver and Sustain. The framework emphasises the behaviours, contextual factors and processes needed to embed effective educational approaches.
A ministry can adapt these into a national Reform Stage-Gate Register:
Explore: identify the need, relevant evidence, alternatives and actual constraints.
Prepare: design the programme, recruit support, confirm legal and technical requirements and define evaluation.
Deliver: implement initially, collect feedback, address emerging barriers and examine whether teaching changes.
Sustain: maintain capability, review impacts and preserve benefits beyond the initial team or funding cycle.
At national scale, add explicit Scale, Repair and Sunset decisions when justified. These are decisions that may be made alongside the four flexible phases, not a claim that every reform must progress through identical fixed dates.
3. Register the theory of change
Every intervention has an implicit explanation of why it should work. A credible programme states that mechanism clearly and tests its important assumptions.
A Policy Theory-of-Change Record should connect inputs, delivery activities, changes to instruction or service, independent learner outcomes and broader public objectives.
For example, a literacy programme may fund teacher training and new lesson resources. Its mechanism might be that teachers more accurately diagnose decoding and comprehension difficulties, select suitable teaching responses and thereby improve independent reading.
The programme should test those links. Teachers attending the workshop is not evidence that diagnostic practice changed. Increased use of a reading app is not evidence that students can understand a new text.
When a link fails, the ministry should know whether to strengthen implementation, revise the programme or question the original policy mechanism.
4. Check for prior evidence and existing national programmes
New policies are attractive, but they may duplicate services already funded elsewhere. A proposed school counselling pilot, for instance, might overlap with established specialist services or use less-qualified staff without clear added benefit.
A Reform Collision and Evidence Check should review current programmes, national standards, prior trials, relevant international studies, existing contracts, licences and legal obligations.
Avoid copying successful international models without examining differences in curriculum, teaching expertise, language and student population.
The OECD Education Policy Outlook 2025
A proposed reform should name the educational problem left unsolved by current provision and explain why another programme is warranted.
5. Define the pilot population fairly
An intervention tested only in highly motivated schools with exceptional leadership may fail when introduced to typical schools. A pilot should be designed around the question the ministry needs to answer.
A Pilot Site Selection Record should explain what populations and conditions are included: school stage, subject, region, language, prior achievement, staff capability, infrastructure and relevant accessibility needs.
Selection may be random or purposefully varied, depending on the evaluation design. The ministry must state the limits of what the chosen sample can establish.
Do not present success in five unusually well-resourced schools as proof of national feasibility. A pilot can establish that an idea is possible under certain conditions without proving that ordinary schools can reproduce it.
Where some groups are excluded from the pilot, explain why and how their needs will be considered before future scale-up.
6. Measure the starting position accurately
A pilot without a credible baseline may be unable to distinguish improvement from selection effects, normal development or changes occurring elsewhere.
A Pilot Baseline Evidence Record should identify the intended student capability, current performance, population, relevant contextual conditions, measurement limits and date.
For a mathematics intervention, use suitable evidence of mathematical understanding rather than rely solely on a general report-card grade. A teacher development project may need an appropriate measure of instructional practice as well as pupil outcomes.
Collect only the necessary personal information under valid privacy and consent arrangements. An elaborate new data collection programme is not justified solely because analysts would find it convenient.
A baseline is a comparison point, not a permanent label on children or schools.
7. Decide what evidence is sufficient for the stage
The ministry needs different levels of confidence for different decisions. A small internal improvement in how schools file records may require a modest quality check. A high-stakes change to national examinations requires far stronger evidence and formal authority.
A Stage-Gate Evidence Standard should reflect risk, cost, legal consequence, target population and reversibility.
The OECD Education Policy Outlook 2024 analysis of monitoring and rapid evaluation notes the practical value of flexible, timely evidence that informs and modifies implementation. Its discussion distinguishes responsive monitoring from long, expensive impact evaluations.
A ministry can use qualitative implementation evidence, school observation, appropriately designed comparisons, independent student work and specialist reviews according to the question.
No single statistical method guarantees a correct decision. Explain what is known and what further evaluation would be needed before making a stronger causal claim.
8. Distinguish fidelity from justified adaptation
A programme may have critical components that must remain consistent if its mechanism is to work. It may also have superficial or contextual features that schools should adapt.
A Core-and-Adaptable Components Map should identify:
Core: essential teaching or service mechanisms supported by credible evidence.
Adaptable: language, examples, delivery schedule or other elements that can change without defeating the mechanism.
Constrained: legal, safety, examination and accessibility requirements that must be followed.
Experimental: untested variations requiring explicit monitoring.
For example, a literacy programme may require a coherent sequence of specific decoding and comprehension instruction while allowing appropriate language examples. A school may adapt scheduling within limits without omitting essential components.
If implementation differs substantially from the intended programme, the evaluation must reflect what schools actually delivered, rather than judge a theoretical model that never existed in practice.
9. Budget the full cost of implementation
Reforms can appear inexpensive when only the software licence or initial training fee is counted. National implementation may also require staff release time, materials, supervision, accommodation, infrastructure, evaluation, maintenance and transition support.
A Reform Total-Cost Ledger should include:
- Design and technical development.
- Programme materials and licensing.
- Teacher and school leader preparation.
- Substitute staffing or release time.
- Classroom or specialist equipment.
- Accessibility and language adaptation.
- Programme coordination and data systems.
- Independent evaluation and quality assurance.
- Support for lower-capacity schools.
- Maintenance, renewal and possible exit costs.
Schools should not be expected to fund the missing portion through unpaid teacher labour or compulsory family purchases.
A programme sustainable in a ten-school pilot may become unaffordable at national scale. Assess that before promising universal rollout.
10. Protect teacher capability as a scale condition
A national reform is implemented through ordinary professionals, not only the experts who wrote the policy. If the new curriculum requires unfamiliar pedagogy or technology, teachers need preparation and time.
A Teacher Implementation Readiness Check should verify subject knowledge, understanding of the new practice, access to model lessons, realistic rehearsal, coaching, materials and the ability to respond to learner difficulties.
One online briefing is unlikely to be sufficient preparation for a major change in instructional practice. Conversely, a familiar administrative change may require only brief, clear communication.
Teacher readiness should be based on practical evidence where appropriate, not the number of certificates distributed.
See Teacher Education, Practicum and Instructional Coaching.
11. Align implementation with school operating conditions
A reform may require laboratories, specialist staff, particular devices or new assessment arrangements. If the resources arrive months later, the policy may fail for reasons unrelated to its central idea.
A School Implementation Dependency Gate should inspect actual facilities, curriculum materials, qualified workforce, timetables, accessibility, transport where relevant, vendor readiness and school leadership capacity.
Allow schools to declare important dependencies unmet through a legitimate escalation route. Do not require principals to falsely report full readiness for fear of penalties.
The central ministry should coordinate decisions rather than allow separate departments to announce that their part is complete while the learner-facing service remains impossible.
12. Evaluate educational outcomes independently of delivery metrics
A programme may distribute tablets, increase student login frequency and train many teachers without producing the intended educational improvement.
A Pilot Outcome Evidence Map should separate participation, actual delivered instruction, demonstration of target capability, learner access, equity, staff workload and longer-term retention or transfer.
For a reading intervention, independent comprehension on unfamiliar text may be more meaningful than app completion. For a vocational initiative, legitimate practical performance matters more than attendance in theory sessions.
Be cautious about attribution. Students may improve as they grow, teachers may alter other practices, and examination format changes can affect scores. The evaluation should distinguish descriptive change from supported causal findings.
The research governance framework is in Education Research and Evidence Synthesis.
13. Conduct a formal scale-readiness review
A positive pilot is not the same as a scalable service. Before expanding to more schools, the ministry should require a National Scale Readiness Certificate or equivalent authorised decision record that covers:
- What the intervention demonstrably achieves.
- Which learners benefit and which may be underserved.
- The level of confidence in the evidence.
- Essential professional and infrastructure conditions.
- Programme costs per realistic unit of service.
- Staffing and training availability.
- Quality and safety obligations.
- Legal and procurement readiness.
- Data and interoperability requirements.
- How implementation will be monitored during growth.
- Contingencies if important assumptions fail.
- Criteria for pausing, adapting or stopping expansion.
The certificate should record approval authority and unresolved conditions. Approval can be conditional, with specified matters that must be resolved before the next stage.
14. Scale in cohorts rather than as an uncontrolled announcement
Large reforms may be implemented by grade, subject, district or school readiness, depending on the educational and legal purpose. Phased expansion gives authorities opportunities to learn and correct.
The OECD Education Policy Outlook 2025
A Scale Wave Plan should identify sites, cohort dates, capacity, materials, support owners, common standards, expected risks and follow-up evidence.
Do not expand faster than the system can supply qualified staff or necessary resources. A national launch date should not become an excuse to turn known problems into compulsory school-level workarounds.
15. Protect schools that were not in the pilot
Schools excluded from the first pilot may have weaker infrastructure, more complex access needs or different student populations. They should not receive a worse version of the programme because the initial successful sites received unusually intensive support.
A Non-Pilot School Readiness Audit should compare staffing, facilities, language access, student needs and ordinary operating capacity with pilot conditions.
The ministry may need differentiated implementation support, additional preparation or a revised delivery method. This is not automatically lowering standards; it can be necessary to protect a common educational floor.
When adaptations are substantial, collect new evidence. An intervention proven under one configuration should not be described as equally established under an entirely different one without further verification.
16. Monitor unintended consequences while scaling
Reforms can improve one indicator while creating problems elsewhere. More frequent assessments may narrow teaching. New technology can shift household costs. A specialised curriculum can crowd out another subject. A teacher incentive can encourage selective intake.
A Reform Side-Effect Register should monitor learning time, equity, access, workload, spending, privacy, student protection, related curricular outcomes and the reliability of dependent services.
The ministry should provide safe reporting from schools and learners and distinguish genuine implementation problems from complaints rooted in ordinary uncertainty about change.
The aim is not to abandon every innovation when something is difficult. It is to identify which problems can be repaired and which reveal a mistaken underlying design.
17. Establish continuation and repair gates
After a defined evaluation period, a reform should reach an explicit decision rather than quietly continue because the original funding or political announcement created momentum.
A Reform Continuation Decision may authorise:
Continue: sufficient evidence of benefit and sustainable delivery.
Improve implementation: a plausible mechanism with repairable delivery failures.
Adapt and retest: changes to key elements that require new evaluation.
Limit scope: benefits confined to particular populations or conditions.
Pause: unresolved safety, legality, quality or capacity concerns.
Stop or sunset: unacceptable risk, repeated failure, lack of educational value or a better alternative.
Every decision should state the evidence and responsible authority. A programme can be promising but not yet suitable for national expansion.
18. Make sunset rules part of the original approval
Stopping a public education programme can be difficult once schools, families, employees and vendors depend on it. That is why sunset conditions should be defined before implementation rather than after political enthusiasm has faded.
A Programme Sunset and Teach-Out Plan should identify:
- The conditions requiring termination or major redesign.
- Authorised decision-maker and review schedule.
- Rights and obligations of enrolled learners.
- Alternative teaching and assessment arrangements.
- Recognition of completed study and qualifications.
- Staff and supplier contract obligations.
- Safe data export, retention and deletion.
- Disposal or reuse of materials and equipment.
- Communication to schools and families.
- A final evaluation of benefits, costs and lessons.
Stopping the intervention must not mean abandoning students midway through a recognised course. Public service continuity is part of responsible withdrawal.
19. Preserve institutional memory from failed trials
A pilot that did not achieve the intended benefit can still teach the ministry something valuable. If its records disappear, future teams may repeat the same intervention under a different name.
A National Reform Learning Ledger should preserve the original problem statement, theory of change, implementation design, participating schools, evidence, limitations, costs, unintended effects and final decision.
The record should distinguish the failure of an idea from inadequate delivery. It should also explain what cannot be concluded from the evidence.
Staff should not be encouraged to conceal weaknesses to protect the reputation of a minister, vendor or school. Properly governed learning from failure is part of institutional strength.
This connects to Ministry Operating Model, Civil Service Capability and Institutional Memory.
20. Prevent reform overload and competing initiatives
A ministry may launch literacy, mathematics, technology, safeguarding, assessment and reporting reforms simultaneously. Even individually sound programmes can fail when they compete for the same teachers, leadership time and school resources.
A National Reform Portfolio Calendar should map initiatives by school stage, required staff, training hours, procurement dependencies, overlapping learners, data demands and likely workload.
When two reforms require the same teachers during the same period, the ministry should resolve the conflict before implementation. It may need to sequence initiatives or retire lower-value tasks.
A school should not be expected to satisfy an unlimited list of centrally announced priorities simply through harder work.
The timing standard is developed in Instructional Time, School Calendars and Learning Load.
21. Case study: a promising algebra pilot expands too quickly
Imagine a mathematics team developing a new approach to connecting fractions with algebra. Ten pilot schools receive expert coaching, reviewed materials and extra time for teachers to study student misconceptions. Their pupils demonstrate better independent reasoning.
The ministry then announces immediate national adoption, but ordinary schools receive only a short online presentation. Teachers lack the worked examples and coaching that made the pilot successful. Student work shows that the new approach has become another procedural worksheet sequence.
A gold-standard implementation system separates programme failure from scaling failure. It identifies the core instructional method, confirms the necessary coaching and resources, and phases expansion through schools that can actually deliver it. Learners are assessed on transfer to unfamiliar algebra problems, not the number of lessons using the new brand.
The ministry revises the timetable and offers suitable support before further rollout. A genuinely useful idea survives because the system respects the conditions needed for it to work.
22. Common failure modes and their repair
The pilot success halo
Failure: unusually supported schools are treated as representative of the entire system. Repair: test readiness under ordinary conditions.
Announcement before diagnosis
Failure: the proposed technology or method defines the problem it claims to solve. Repair: register the educational need and plausible alternatives first.
Training completion as success
Failure: certificates replace evidence of changed teaching. Repair: inspect practice and independent learner outcomes.
Scaling without capacity
Failure: rollout begins before staffing, facilities or materials are ready. Repair: require realistic implementation gates.
Outcomes without attribution
Failure: every positive trend during a pilot is credited to the reform. Repair: use suitable comparative evaluation and proportionate confidence.
Permanent pilots with no decision
Failure: short-term programmes continue indefinitely without benefit review. Repair: set continuation and sunset decisions.
Withdrawal that harms current learners
Failure: stopping a programme leaves students without a recognised route. Repair: fund appropriate teach-out and transfer arrangements.
23. The first 180 days
Days 1–30: inventory active reforms and pilots, their stated problems, costs, evidence, owners and commitments. Identify urgent conflicts or unacceptable risks.
Days 31–60: establish a Reform Stage-Gate Register and theory-of-change requirements. Examine a sample of pilots for baseline and implementation quality.
Days 61–90: define a scale-readiness decision standard and conditions for programme continuation or pause.
Days 91–120: pilot the gate on one intervention nearing expansion and on one programme whose educational value is uncertain.
Days 121–180: publish authorised decisions, verified learning and financial consequences, including justified adaptation, continuation or retirement.
24. Gold-standard acceptance tests
- Does each major reform identify a real educational problem and credible alternatives?
- Is its mechanism explained through a testable theory of change?
- Are pilot populations and conditions sufficiently transparent to interpret results?
- Are learning effects separated from mere activity or software engagement?
- Can the ministry distinguish core programme elements from acceptable adaptation?
- Are full implementation costs, staff time and support needs modelled?
- Does national scale-up require a specific evidence and readiness decision?
- Are unintended effects on equity, curriculum and workload monitored?
- Does every major pilot have a continuation, correction or sunset rule?
- Can students and teachers rely on a lawful transition if a programme is withdrawn?
25. Reference sources and related chapters
The principal resources are EEF, A School’s Guide to Implementation (Third Edition, 2024), EEF Review of Evidence on Implementation in Education (2024), OECD, Education Policy Outlook 2024: Strengthening Capacity in Evaluation and Monitoring and OECD Education Policy Outlook 2025.
This is Chapter 74 of The Gold Standard of a Ministry of Education Reference Manual. It complements Strategic Foresight and Policy Renewal, Evidence Synthesis and Policy Learning, National Performance and Public Reporting and National Risk and Assurance.
Final principle
The gold standard of education reform is not launching the most new programmes. It is knowing which problems deserve intervention, which approaches improve real learning, what conditions make them dependable at scale and when a public institution should have the courage to adapt or stop something that does not work.